Compare commits

...
Author SHA1 Message Date
Abdul WahabandCursor 6bea544a0a Fix: drain context stdout before process.exit (#573)
process.exit after a queued write truncated boot output on a backpressured pipe. Await the write callback, then close the fetch dispatcher.

AI assistance: implemented with Cursor Grok 4.6.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-25 07:51:47 +05:00
Abdul WahabandCursor 2ef8e43d1e Fix: close fetch sockets before context helper exit (#573)
On Windows/Node 24, a live undici keep-alive from the update-check fetch aborted libuv during teardown after valid stdout. Destroy the dispatcher first, matching concept-seed.

AI assistance: implemented with Cursor Grok 4.6.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-25 07:31:18 +05:00
dependabot[bot]andGitHub 78b50aa416 Build(deps): update the Bun dependency group (#650)
Update the grouped Bun dependencies while holding ai at 7.0.69 to avoid the automatic tool-execution behavior regression introduced in 7.0.70.

Prepared and validated with AI assistance.
2026-08-24 15:15:01 -04:00
Abdul WahabandGitHub c3a30086bc Merge pull request #649 from pbakaus/codex/link-hook-trust-guide
Docs: Link harness trust guidance
2026-08-24 08:26:26 +05:00
Abdul Wahab 21510c3632 Docs: Link harness trust guidance
Point installer readers to the harness-specific approval and verification steps on impeccable.style.\n\nAI-assisted: Codex prepared and verified this documentation update under direct maintainer instruction.
2026-08-24 08:00:38 +05:00
github-actions[bot] 5d00e30405 Sync generated provider output 2026-08-24 02:52:15 +00:00
Abdul WahabandGitHub f01a808890 Merge pull request #647 from pbakaus/fix/603-codex-stop-payload
Fix: emit Codex Stop hook as decision/block (#603)
2026-08-24 07:51:42 +05:00
Abdul WahabandCursor 2064b0696f Merge origin/main into fix/603-codex-stop-payload
Keep Codex as its own Stop harness (decision/block) while taking main's Grok envelope detection and Stop cache sync.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-24 07:25:47 +05:00
github-actions[bot] c3ea1c79f5 Sync generated provider output 2026-08-24 02:18:59 +00:00
Abdul WahabandGitHub f849d610f3 Merge pull request #648 from pbakaus/fix/646-grok-hook-stdin
Fix: parse Grok Build camelCase hook stdin (#646)
2026-08-24 07:18:24 +05:00
Abdul WahabandGitHub c0b1ec6fef Merge pull request #644 from pbakaus/fix/642-grok-global-hook
Fix: rewrite Grok project hooks to the global skill path (#642)
2026-08-24 06:14:38 +05:00
Abdul WahabandCursor c9e7cd8a64 Fix: emit Codex Stop hook as decision/block (#603)
Codex Stop rejects Claude's hookSpecificOutput shape. Detect Codex from
turn_id at runtime and emit { decision: "block", reason } so existing
installs keep working without rewriting hook commands.

AI-assisted change, prepared with Cursor Grok under maintainer direction.

Fixes #603
Fixes #643

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-24 05:23:51 +05:00
Abdul WahabandCursor 49571365a8 Fix: rewrite Grok project hooks to the global skill path (#642)
Grok was skipped by the hook-command rewrite, so a global skill install left .grok/hooks/impeccable.json pointing at a project-relative hook.mjs that does not exist.

AI assistance: Cursor Grok 4.6 implemented this change.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-24 04:27:01 +05:00
58 changed files with 2523 additions and 654 deletions
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook — PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook — PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+2
View File
@@ -113,6 +113,8 @@ npx impeccable update
Codex users should open `/hooks` after install or update and approve the project hook when prompted. Codex tracks trust by hook definition, so updates that change `.codex/hooks.json` can require approval again. Grok Build users need project folder trust (`/hooks-trust` or launch with `--trust`) before `.grok/hooks/` scripts run.
See [Allow the hook in your harness](https://impeccable.style/docs/hooks#allow-the-hook-in-your-harness) for harness-specific trust and verification steps.
### Option 2: Git Submodule
For teams that want to keep Impeccable vendored and updated through Git, add this repo as a submodule and link the compiled provider build into your harness folders:
+23 -19
View File
@@ -17,7 +17,7 @@
"@ai-sdk/google": "^4.0.8",
"@ai-sdk/openai": "^4.0.7",
"@anthropic-ai/claude-agent-sdk": "^0.3.165",
"@anthropic-ai/sdk": "^0.115.0",
"@anthropic-ai/sdk": "^0.120.0",
"@babel/parser": "^8.0.4",
"ai": "^7.0.14",
"archiver": "^8.0.0",
@@ -31,37 +31,37 @@
},
},
"packages": {
"@ai-sdk/anthropic": ["@ai-sdk/anthropic@4.0.39", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-JAMGtYeEuaBzqbsPO4fkho6vQyNoVhsHASM4o59wmJRU6Vh7prjOp490Kmc7YQTY+ioU1/xYzXvWOtxZBup0Xw=="],
"@ai-sdk/anthropic": ["@ai-sdk/anthropic@4.0.40", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.28" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-cFlCZspCUC1LGDipARsKx3+4A8c9qI+vFuG0/04Phs0deKwifNsl8wDmcU2HO31aiNR9AJgEBcvg5S00zUS70g=="],
"@ai-sdk/gateway": ["@ai-sdk/gateway@4.0.52", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27", "@vercel/oidc": "3.2.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-SXUM8jzzuTUJRq+EOgPd5to6DSx0EKslVn+IVZHbUEX6k/3vCPNrvjckbK26HnNxHU/STxm+zTSJteqrO+7Z0w=="],
"@ai-sdk/gateway": ["@ai-sdk/gateway@4.0.55", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27", "@vercel/oidc": "3.2.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-7WP/nlDz2BkXFlZzwF3w5JgCvktyHC++LP4PZ4mQpbvxrq+M7OdNslevlkuddHhJ+62BZu4oiL/afXlsWOSJZQ=="],
"@ai-sdk/google": ["@ai-sdk/google@4.0.44", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-bmRTDg06jQD+eX8nf214pET9+Oe8O1+lUIRGbWsGXj9IN2UJkpl1O1x7cvtiboyTtKSLvSRdVtItUfSl8sQ2GA=="],
"@ai-sdk/google": ["@ai-sdk/google@4.0.49", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.28" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-ZJMehjQNVyvLytV5cUQcTMNEG79NS+p8d2AHu+bH/Mg2M2IQ8Pueh/xQga2v85TX8BGT55m9Q01taY1fZjS4ag=="],
"@ai-sdk/openai": ["@ai-sdk/openai@4.0.42", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-ZxDca6jJalYuXrIGVrw6dnkpz1Io9AWy+/b/wVWIbjigHCbd+zWLpPi8NnK0OFU+U3YCpP+KWfUvEnG5pFhltA=="],
"@ai-sdk/openai": ["@ai-sdk/openai@4.0.45", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.28" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-F/zHXTfCaHO0Q5egf8HRdc645Mp/ofLE9NSjl5RwaXfqrqLfiLROE9xScgKf4iziElsjuQYbesPEcW37mmZy3w=="],
"@ai-sdk/provider": ["@ai-sdk/provider@4.0.7", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-6or44XprPzKbr8zkmzosowSE0pxkvJcoojBL+mCZvPUt3kvXp3XSNqeVun9golb1acEfSo6yaEBRT18h2VU+1Q=="],
"@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@5.0.27", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@standard-schema/spec": "^1.1.0", "@workflow/serde": "4.1.0", "eventsource-parser": "^3.0.8", "undici": "^7.28.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-EzAn4pdgG5g0xXtH6lE2zyNmfjDQIDjATkfqzuidEI35g++hh4+07vnjzkT/RmGmIClPZiRj/Q2GMPV2V7mkHw=="],
"@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@5.0.28", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@standard-schema/spec": "^1.1.0", "@workflow/serde": "4.1.0", "eventsource-parser": "^3.0.8", "undici": "^7.28.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-TnHUyd/rCYQqHg5RuiOaz/hUql6U+kbUaBW0Rp+0N5UhnAInA9CzzV0HXvuAAPwppDsK6fAx9Rd+tRawpJ/3pg=="],
"@anthropic-ai/claude-agent-sdk": ["@anthropic-ai/claude-agent-sdk@0.3.233", "", { "optionalDependencies": { "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.233", "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.233", "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.233", "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.233", "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.233", "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.233", "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.233", "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.233" }, "peerDependencies": { "@anthropic-ai/sdk": ">=0.93.0", "@modelcontextprotocol/sdk": "^1.29.0", "zod": "^4.0.0" } }, "sha512-Dy+YqhggwtbezDy3Ap2pb1sK3bOqnI+sLNnsVjB3AUWvR0QlGnjjrjORXY03Y50I+B1eFRNEcYPAZKRYlCkSLQ=="],
"@anthropic-ai/claude-agent-sdk": ["@anthropic-ai/claude-agent-sdk@0.3.238", "", { "optionalDependencies": { "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.238", "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.238", "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.238", "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.238", "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.238", "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.238", "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.238", "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.238" }, "peerDependencies": { "@anthropic-ai/sdk": ">=0.93.0", "@modelcontextprotocol/sdk": "^1.29.0", "zod": "^4.0.0" } }, "sha512-ppRfbAflZuV7HPqr8BkHPEk8c7gih3PK+GHnZq6zP0Uo7bJHXAJwc4Za8LJRm4hXMixBNMjK1cU/VWfQ9fE2sg=="],
"@anthropic-ai/claude-agent-sdk-darwin-arm64": ["@anthropic-ai/claude-agent-sdk-darwin-arm64@0.3.233", "", { "os": "darwin", "cpu": "arm64" }, "sha512-4WDiBZgcrmvTDJjS8RNZwoxGgMz/0EpOM+sYa6EtyjwHTd6It1H/+k5zBckCmBajbgS5/ASCJqdwZzi7dwBl0Q=="],
"@anthropic-ai/claude-agent-sdk-darwin-arm64": ["@anthropic-ai/claude-agent-sdk-darwin-arm64@0.3.238", "", { "os": "darwin", "cpu": "arm64" }, "sha512-7KctNItTzHRiDg+jFxTM46o+Y/YS52V6HSopaWPggO6BbR+3MV/rKMmq+zP93If7eVwadW9Yg9vHLjXDKIw9OQ=="],
"@anthropic-ai/claude-agent-sdk-darwin-x64": ["@anthropic-ai/claude-agent-sdk-darwin-x64@0.3.233", "", { "os": "darwin", "cpu": "x64" }, "sha512-RaaEfNrbqSh77H5NdVF9cJQ0xhAUO92aOv71LSKSdAYModMeUvJN0k22Q7gvmx0TlmqJ+aVyCG8J8gVfgSL9mg=="],
"@anthropic-ai/claude-agent-sdk-darwin-x64": ["@anthropic-ai/claude-agent-sdk-darwin-x64@0.3.238", "", { "os": "darwin", "cpu": "x64" }, "sha512-aDt8LXjWISwLzqeCBHxJC5AR1iVYlY5UYIm2VLztjI1U0PCb1BiH/IfyyjjizqRhhEwyQp9YAJeZZq8n/2vQEA=="],
"@anthropic-ai/claude-agent-sdk-linux-arm64": ["@anthropic-ai/claude-agent-sdk-linux-arm64@0.3.233", "", { "os": "linux", "cpu": "arm64" }, "sha512-Az9HjQthYQqRjJCacBtDIAHX3TRGK9WlACNb/UOGAK3JndNzZMprM2mK/t6YmP2cRLJsGyorxL7HZmR9R9HYaw=="],
"@anthropic-ai/claude-agent-sdk-linux-arm64": ["@anthropic-ai/claude-agent-sdk-linux-arm64@0.3.238", "", { "os": "linux", "cpu": "arm64" }, "sha512-P7V9TFokcNRdIJPUzDQ0GLBfMuoewWEH4rh6U3e7RaEAxzdeUcdS9P0N4arXqK2YAOtK1o93QSEtiGUF1fTfWw=="],
"@anthropic-ai/claude-agent-sdk-linux-arm64-musl": ["@anthropic-ai/claude-agent-sdk-linux-arm64-musl@0.3.233", "", { "os": "linux", "cpu": "arm64" }, "sha512-Z3uZdzt6xgJ3f4NIgO6lzBYSELULKSq6AL4OsNLBzuaEpVW0iYs1kUCaD9rcMlMrf3cV+Dk/GA/lTCGMgbucjQ=="],
"@anthropic-ai/claude-agent-sdk-linux-arm64-musl": ["@anthropic-ai/claude-agent-sdk-linux-arm64-musl@0.3.238", "", { "os": "linux", "cpu": "arm64" }, "sha512-ankSEMAMTVulKYg0NT8fZ7a5+q+aIzfiqhcrLe9zYPeuhE+lpNeq0JPMcHyOtUAENxVI0/2J2q5OXGl9O0qWcg=="],
"@anthropic-ai/claude-agent-sdk-linux-x64": ["@anthropic-ai/claude-agent-sdk-linux-x64@0.3.233", "", { "os": "linux", "cpu": "x64" }, "sha512-jpbhV+n9PnxLiyheQ/HjtHIg/E5/jVsk2Vdu132BSoL/3bsObSmMqKgsqoMutzwRZvtpqRs2RPVcjsC8G4A9Zw=="],
"@anthropic-ai/claude-agent-sdk-linux-x64": ["@anthropic-ai/claude-agent-sdk-linux-x64@0.3.238", "", { "os": "linux", "cpu": "x64" }, "sha512-/arAOSqtIAWDu7Z4Uf2Un3n+/5Zg3NpxZKzMQEDnYNkbJRhr44sM59TFaEOiWq1FqlUJKi6wHjgm+TW5B8DWWg=="],
"@anthropic-ai/claude-agent-sdk-linux-x64-musl": ["@anthropic-ai/claude-agent-sdk-linux-x64-musl@0.3.233", "", { "os": "linux", "cpu": "x64" }, "sha512-kYBIAQCu2f1YITcGbpUN2jfrkAzs59TVAragAhE2z+GrkIcxcpZwmaRY6heMBtaSY8SuyrwgqbCW9hJALYFnEg=="],
"@anthropic-ai/claude-agent-sdk-linux-x64-musl": ["@anthropic-ai/claude-agent-sdk-linux-x64-musl@0.3.238", "", { "os": "linux", "cpu": "x64" }, "sha512-buo3IBSd7EmYcQsH+OKARgNNbSF1+l/+z1hDXSXGKzwD1LFioJZ/k9FWWWC166SoQ93jPmN+G/ksTzEUB3zRGQ=="],
"@anthropic-ai/claude-agent-sdk-win32-arm64": ["@anthropic-ai/claude-agent-sdk-win32-arm64@0.3.233", "", { "os": "win32", "cpu": "arm64" }, "sha512-aO2MaNdmQofyPLKszE4s+Ope/sLJPeI/ZlGdCcjYp7qhji2hgZ4bRWWsOrx5eKjz0gFK5CFFltILkFcNcxCsVg=="],
"@anthropic-ai/claude-agent-sdk-win32-arm64": ["@anthropic-ai/claude-agent-sdk-win32-arm64@0.3.238", "", { "os": "win32", "cpu": "arm64" }, "sha512-yW8lhV7QgYiNu3NatjgmkhUspgJsG2N2N6lmm/7B99sWFobKTbVLMLsXxj6A46R6qE2C4Zzm6PjxgGaMYOMn0g=="],
"@anthropic-ai/claude-agent-sdk-win32-x64": ["@anthropic-ai/claude-agent-sdk-win32-x64@0.3.233", "", { "os": "win32", "cpu": "x64" }, "sha512-TcAYyWPXS5mREZGUksuCZsLIRQjbo/Vriur2PqIhAmgZ1oiqBZO27a90sX60EUczD7yV8wpwOVhVLhUxO0kAEg=="],
"@anthropic-ai/claude-agent-sdk-win32-x64": ["@anthropic-ai/claude-agent-sdk-win32-x64@0.3.238", "", { "os": "win32", "cpu": "x64" }, "sha512-6Fb2JRrBci282fBfhCGB0Tpq7N8V5HMMqsO3YGFkc9hhbn/y9G7glNKSNgRtxD5Ujjxj8rhiaYstXW2DJzanzw=="],
"@anthropic-ai/sdk": ["@anthropic-ai/sdk@0.115.0", "", { "dependencies": { "json-schema-to-ts": "^3.1.1", "standardwebhooks": "^1.0.0" }, "peerDependencies": { "zod": "^3.25.0 || ^4.0.0" }, "optionalPeers": ["zod"], "bin": { "anthropic-ai-sdk": "bin/cli" } }, "sha512-BJrFIVyjNuU8lfDyIJTvlRYzgQg+zEl78BxE7fq8esULsGz9IRQvGtW5spq3tydmtjQb/GFdooKGdGsetpx+lQ=="],
"@anthropic-ai/sdk": ["@anthropic-ai/sdk@0.120.0", "", { "dependencies": { "json-schema-to-ts": "^3.1.1", "standardwebhooks": "^1.0.0" }, "peerDependencies": { "zod": "^3.25.0 || ^4.0.0" }, "optionalPeers": ["zod"], "bin": { "anthropic-ai-sdk": "bin/cli" } }, "sha512-ZlvmNFT/iIF6JD13rxbbMWD8nvGR0RaUp6yMQnoc+4Af0YjVVe/bIdW1XSQQsoxXAtg1NaT6Vak0LKFlJ4d37Q=="],
"@babel/helper-string-parser": ["@babel/helper-string-parser@8.0.0", "", {}, "sha512-6mJgmFFFIIO82vvoLt9XtRC7/TkzXfts1t/SpRX4IHSzMgqoPYCWesVu1udUPUWioAE/2fcG6WuI8zrkE1gwrg=="],
@@ -109,7 +109,7 @@
"acorn": ["acorn@8.17.0", "", { "bin": { "acorn": "bin/acorn" } }, "sha512-xRQbDb9BnwDafYNn6Vwl839DYVjqXYb1XVGtWAZ1kcDc6iwAL4hg3B1dZlRiuENFeO2H53gFG3in621AdERVAg=="],
"ai": ["ai@7.0.66", "", { "dependencies": { "@ai-sdk/gateway": "4.0.52", "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-wBUyoCYF3GVr+62nelBgR8YbpTSsMZrzFyOOjiwijylNSM2TFCW35C+Pml2vc59/WLMpyhS/LWZ55M+B9DAcSg=="],
"ai": ["ai@7.0.69", "", { "dependencies": { "@ai-sdk/gateway": "4.0.55", "@ai-sdk/provider": "4.0.7", "@ai-sdk/provider-utils": "5.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-gudEqQYt/FuRpQkckLHPYKOh5M5v5ayPmzQAk5xyZOaWkUM0ArNGdXQiBGEuZA7GCwYOjzmWxB44ZW0CyRhHfQ=="],
"ajv": ["ajv@8.18.0", "", { "dependencies": { "fast-deep-equal": "^3.1.3", "fast-uri": "^3.0.1", "json-schema-traverse": "^1.0.0", "require-from-string": "^2.0.2" } }, "sha512-PlXPeEWMXMZ7sPYOHqmDyCJzcfNrUr3fGNKtezX14ykXOEIvyK81d+qydx89KY5O71FKMPaQ2vBfBFI5NHR63A=="],
@@ -327,7 +327,7 @@
"magic-string": ["magic-string@0.30.21", "", { "dependencies": { "@jridgewell/sourcemap-codec": "^1.5.5" } }, "sha512-vd2F4YUyEXKGcLHoq+TEyCjxueSeHnFxyyjNp80yg0XV4vUhnDer/lvvlqM/arB5bXQN5K2/3oinyCRyx8T2CQ=="],
"marked": ["marked@18.0.9", "", { "bin": { "marked": "bin/marked.js" } }, "sha512-/Sa4qiiHZxf0/FQdBBowr9q4r10krCwMvpK48FUBdXdUXScDxiQGR9zCPrFgRVR5LU3iySOiIjy09ZQvADir1w=="],
"marked": ["marked@18.0.10", "", { "bin": { "marked": "bin/marked.js" } }, "sha512-FJeH4bRpYoXiggcgriCGItKCSv3xkngJc4QCZ/rkQCogU3VYaLxYJoZl8Nw/b4+x7iij/pd+09mZ6A1dXzpL0A=="],
"math-intrinsics": ["math-intrinsics@1.1.0", "", {}, "sha512-/IXtbwEk5HTPyEwyKX6hGkYXxM9nbj64B+ilVJnC/R6B0pH5G4V3b0pVbL7DBj4tkhBAppbQUlf6F6Xl9LHu1g=="],
@@ -435,7 +435,7 @@
"strip-ansi": ["strip-ansi@7.2.0", "", { "dependencies": { "ansi-regex": "^6.2.2" } }, "sha512-yDPMNjp4WyfYBkHnjIRLfca1i6KMyGCtsVgoKe/z1+6vukgaENdgGBZt+ZmKPc4gavvEZ5OgHfHdrazhgNyG7w=="],
"svelte": ["svelte@5.56.9", "", { "dependencies": { "@jridgewell/remapping": "^2.3.4", "@jridgewell/sourcemap-codec": "^1.5.0", "@sveltejs/acorn-typescript": "^1.0.10", "@types/estree": "^1.0.5", "@types/trusted-types": "^2.0.7", "acorn": "^8.12.1", "aria-query": "5.3.1", "axobject-query": "^4.1.0", "clsx": "^2.1.1", "devalue": "^5.8.1", "esm-env": "^1.2.1", "esrap": "^2.2.12", "is-reference": "^3.0.3", "locate-character": "^3.0.0", "magic-string": "^0.30.11", "zimmerframe": "^1.1.2" } }, "sha512-VT8kSnlEg8069w7AiCcAk3Yf5xvMnrGTagVOmU/OpOLHaHnNqXhWZCH/4EVga/bT/HtWhvE6/fHrXLErx7OnJA=="],
"svelte": ["svelte@5.56.10", "", { "dependencies": { "@jridgewell/remapping": "^2.3.4", "@jridgewell/sourcemap-codec": "^1.5.0", "@sveltejs/acorn-typescript": "^1.0.10", "@types/estree": "^1.0.5", "@types/trusted-types": "^2.0.7", "acorn": "^8.12.1", "aria-query": "5.3.1", "axobject-query": "^4.1.0", "clsx": "^2.1.1", "devalue": "^5.8.1", "esm-env": "^1.2.1", "esrap": "^2.2.12", "is-reference": "^3.0.3", "locate-character": "^3.0.0", "magic-string": "^0.30.11", "zimmerframe": "^1.1.2" } }, "sha512-Lcxbj8I/KAbpY+VjtY4ENQBV0dDCipfGAhqb51XQZ67CIQqXgsv/8dPkbILaj4Fb6/b6JAEM/PIVbILXgDQy2g=="],
"tar-stream": ["tar-stream@3.1.8", "", { "dependencies": { "b4a": "^1.6.4", "bare-fs": "^4.5.5", "fast-fifo": "^1.2.0", "streamx": "^2.15.0" } }, "sha512-U6QpVRyCGHva435KoNWy9PRoi2IFYCgtEhq9nmrPPpbRacPs9IH4aJ3gbrFC8dPcXvdSZ4XXfXT5Fshbp2MtlQ=="],
@@ -483,10 +483,14 @@
"zod-to-json-schema": ["zod-to-json-schema@3.25.2", "", { "peerDependencies": { "zod": "^3.25.28 || ^4" } }, "sha512-O/PgfnpT1xKSDeQYSCfRI5Gy3hPf91mKVDuYLUHZJMiDFptvP41MSnWofm8dnCm0256ZNfZIM7DSzuSMAFnjHA=="],
"@ai-sdk/gateway/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@5.0.27", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@standard-schema/spec": "^1.1.0", "@workflow/serde": "4.1.0", "eventsource-parser": "^3.0.8", "undici": "^7.28.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-EzAn4pdgG5g0xXtH6lE2zyNmfjDQIDjATkfqzuidEI35g++hh4+07vnjzkT/RmGmIClPZiRj/Q2GMPV2V7mkHw=="],
"@modelcontextprotocol/sdk/eventsource-parser": ["eventsource-parser@3.0.6", "", {}, "sha512-Vo1ab+QXPzZ4tCa8SwIHJFaSzy4R6SHf7BY79rFBDf0idraZWAkYrDjDj8uWaSm3S2TK+hJ7/t1CEmZ7jXw+pg=="],
"@modelcontextprotocol/sdk/zod": ["zod@4.3.6", "", {}, "sha512-rftlrkhHZOcjDwkGlnUtZZkvaPHCsDATp4pGpuOOMDaTdDDXF91wuVDJoWoPsKX/3YPQ5fHuF3STjcYyKr+Qhg=="],
"ai/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@5.0.27", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@standard-schema/spec": "^1.1.0", "@workflow/serde": "4.1.0", "eventsource-parser": "^3.0.8", "undici": "^7.28.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-EzAn4pdgG5g0xXtH6lE2zyNmfjDQIDjATkfqzuidEI35g++hh4+07vnjzkT/RmGmIClPZiRj/Q2GMPV2V7mkHw=="],
"chromium-bidi/zod": ["zod@3.25.76", "", {}, "sha512-gzUt/qt81nXsFGKIFcC3YnfEAx5NkunCfnDlvuBSSFS02bcXu4Lmea0AFIUwbLWxWPx3d9p8S5QoaujKcNQxcQ=="],
"eventsource/eventsource-parser": ["eventsource-parser@3.0.6", "", {}, "sha512-Vo1ab+QXPzZ4tCa8SwIHJFaSzy4R6SHf7BY79rFBDf0idraZWAkYrDjDj8uWaSm3S2TK+hJ7/t1CEmZ7jXw+pg=="],
+3 -3
View File
@@ -1402,7 +1402,7 @@ function hookScriptPathForProvider(skillRoot, provider) {
if (provider === '.cursor') {
return join(skillRoot, provider, 'skills', 'impeccable', 'scripts', 'hook-before-edit.mjs');
}
if (provider === '.claude' || provider === '.agents') {
if (provider === '.claude' || provider === '.agents' || provider === '.grok') {
return join(skillRoot, provider, 'skills', 'impeccable', 'scripts', 'hook.mjs');
}
return null;
@@ -1478,8 +1478,8 @@ function guardHookCommand(quotedPath, provider) {
// entries additionally get a `commandWindows` sibling for cmd.exe.
function rewriteHookCommandsForSkillRoot(value, provider, { skillRoot, absolute }) {
const hookScript = hookScriptPathForProvider(skillRoot, provider);
// Providers we don't own a `node "PATH"` command hook for (.github, .grok)
// carry their own portable command forms; leave them untouched.
// Providers we don't own a `node "PATH"` command hook for (.github) carry
// their own portable command forms; leave them untouched.
if (!hookScript) return value;
// Project-scope installs derive the provider's own project-relative path
+1 -1
View File
@@ -89,7 +89,7 @@
"@ai-sdk/google": "^4.0.8",
"@ai-sdk/openai": "^4.0.7",
"@anthropic-ai/claude-agent-sdk": "^0.3.165",
"@anthropic-ai/sdk": "^0.115.0",
"@anthropic-ai/sdk": "^0.120.0",
"@babel/parser": "^8.0.4",
"ai": "^7.0.14",
"archiver": "^8.0.0",
+3 -3
View File
@@ -2,9 +2,9 @@
Manage the **design detector hook** for the current project.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write.
The hook runs the impeccable design detector on direct file edits to design-relevant files (`.tsx`, `.jsx`, `.html`, `.vue`, `.svelte`, `.astro`, `.css`, `.scss`, `.sass`, `.less`, `.ts`, `.js`). Claude Code, Codex, and GitHub Copilot use a post-tool-use hook and push a short system reminder into the agent's context after the edit; findings get a correction prompt, pending issues get a re-nudge, and clean UI-ish files get a short ack unless quiet mode is on (`hook.quiet` in config). Plain `.ts` and `.js` files are still scanned, but stay quiet unless the detector finds something. Cursor uses `preToolUse` to block bad proposed writes before they land and stays silent when it allows a clean write. Grok Build fires the same PostToolUse scan to mark touched files, then surfaces findings on Stop `additionalContext`. Do not expect a Grok per-edit reminder: Grok discards that stdout.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code and Codex, which both dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit.
The detector rules run in two tiers. The per-edit hook surfaces only the immediate tier: mechanical, unambiguous problems worth interrupting an edit for, such as broken images, overflowing or clipped content, contrast and legibility failures, gradient text, glow shadows, and design-system drift. Everything else (copy cadence, palette and typography taste, layout rhythm) is deferred to a deep pass on the `Stop` hook event, which runs the full rule set over every UI file touched in the session and surfaces the remaining findings once, deduplicated against what the per-edit pass already reported. A session with nothing left to report stops silently. Set `hook.perEditRules` to `"all"` in `.impeccable/config.json` to restore the full rule set on every edit. The Stop deep pass is wired for Claude Code, Codex, and Grok Build, which dispatch a native `Stop` hook event. Cursor does not get one (its stop hook is not consistently dispatched; the pre-write gate covers it), and GitHub Copilot's stop-style events do not feed context back to the model, so they keep the full detector per edit. Grok also fires an observe-only Stop with `reason: "shutdown"` after `end_turn`; skip that one, scan only `end_turn`.
Every hook is a mechanical pass. The reflexes no scanner catches live in [craft-floor.md](craft-floor.md), which the skill loads before it edits UI, so they apply whether or not a hook is wired. A session with no automatic hook gets one `MANUAL_DETECTOR_REQUIRED` directive from `context.mjs` asking for a single detector run at the end.
@@ -14,7 +14,7 @@ Declare server-side template extensions under **`detector.extensions`** when the
Manual `npx impeccable detect` scans use the same project filter config by default: `detector.ignoreRules`, `detector.ignoreFiles`, `detector.ignoreValues`, and `detector.designSystem.enabled`. `hook.enabled` only controls automatic hook execution, not manual CLI scans. Use `npx impeccable detect --no-config ...` for a raw detector run that ignores project config/context. Use `npx impeccable ignores ...` for direct CLI CRUD on the same detector ignores.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
Supported harnesses: Claude Code (`.claude/settings.local.json` in the project, which is gitignored so the hook stays machine-local; a hook you move into the shared `settings.json` is honored in place too), Codex (`.codex/hooks.json` in the project), Cursor (`.cursor/hooks.json` in the project), Grok Build (`.grok/hooks/impeccable.json` in the project; requires `/hooks-trust` or `--trust`), and GitHub Copilot (`.github/hooks/impeccable.json` in the project, a team-shared committed file that both the Copilot CLI and the cloud agent read). For the Copilot CLI, repo-level hooks fire once `.github/hooks/impeccable.json` is committed to the repository's default branch.
On **Cursor**, `preToolUse` checks proposed Write/Edit/Shell write content and denies only when the real detector finds an issue. The denial message is visible to the agent as the tool error, so the agent can reconsider before the bad write lands.
+130 -26
View File
@@ -816,9 +816,9 @@ export function splitFindingsByTier(findings) {
}
// Whether the per-edit pass for this harness should defer non-immediate
// findings to a Stop deep pass. Only Claude Code and Codex dispatch our Stop
// hook; Cursor and GitHub Copilot have no deep pass wired, so deferring for
// them would silently drop the non-immediate rules entirely.
// findings to a Stop deep pass. Claude Code, Codex, and Grok Build dispatch
// our Stop hook; Cursor and GitHub Copilot have no deep pass wired, so
// deferring for them would silently drop the non-immediate rules entirely.
export function perEditTieringActive(config, harness) {
if (harness === 'cursor' || harness === 'github') return false;
return (config?.perEditRules || DEFAULT_CONFIG.perEditRules) !== 'all';
@@ -1251,18 +1251,50 @@ export function resolveHarness(env = {}, event = null) {
const explicit = env?.IMPECCABLE_HOOK_HARNESS;
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
// GitHub Copilot's postToolUse event uses camelCase `toolName`/`toolArgs` and
// has no `tool_name`/`tool_input`. That shape is the discriminator.
if (explicit === 'grok') return 'grok';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
// `tool_input`) also matches Grok, which is how live PostToolUse was
// classified as Copilot and then skipped with no-file-path (#646).
if (looksLikeGrokEnvelope(event)) return 'grok';
if (event && typeof event === 'object'
&& (typeof event.toolName === 'string' || event.toolArgs !== undefined)
&& event.tool_name === undefined && event.tool_input === undefined) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
function looksLikeGrokEnvelope(event) {
if (!event || typeof event !== 'object') return false;
if (event.hook_event_name !== undefined
|| event.tool_name !== undefined
|| event.tool_input !== undefined) {
return false;
}
if (event.toolArgs !== undefined) return false;
if (typeof event.hookEventName === 'string') return true;
return typeof event.toolName === 'string' && event.toolInput !== undefined;
}
// Stop arrives as Claude's `hook_event_name: "Stop"` or Grok Build's
// `hookEventName: "stop"`. hook.mjs routes on the raw stdin, before any
// normalize, so both casings must match here.
export function isStopEvent(event) {
if (!event || typeof event !== 'object') return false;
const name = event.hook_event_name || event.hookEventName;
return typeof name === 'string' && name.toLowerCase() === 'stop';
}
// GitHub Copilot's postToolUse payload is
// { sessionId, timestamp, cwd, toolName, toolArgs, toolResult }
// mapped onto the internal `{ tool_name, tool_input, cwd, session_id }` shape.
@@ -1354,9 +1386,36 @@ function normalizeGitHubEvent(event, projectCwd) {
};
}
// Grok Build 1.0.5 (captured 2026-08-24) sends camelCase `toolName` /
// `toolInput` / `sessionId` / `stopHookActive`, plus `cwd` alongside a
// trailing-slashed `workspaceRoot` (every consumer path.resolve()s, so no
// stripping here). Only the fields the hook reads are copied; the event
// name stays camelCase because routing already happened on the raw stdin
// (isStopEvent) and nothing downstream reads `hook_event_name`.
function normalizeGrokEvent(event, projectCwd) {
const cwd = event.cwd || event.workspaceRoot || envProjectDir(projectCwd) || projectCwd;
const sessionId = event.sessionId || event.session_id || 'unknown';
const rawInput = event.toolInput ?? event.tool_input;
const toolInput = rawInput && typeof rawInput === 'object' && !Array.isArray(rawInput)
? { ...rawInput }
: {};
const out = {
...event,
cwd,
session_id: sessionId,
tool_name: event.toolName || event.tool_name || null,
tool_input: toolInput,
};
if (event.stopHookActive !== undefined && event.stop_hook_active === undefined) {
out.stop_hook_active = event.stopHookActive;
}
return out;
}
export function normalizeHookEvent(event, projectCwd, harness = 'claude') {
if (!event || typeof event !== 'object') return event;
if (harness === 'github') return normalizeGitHubEvent(event, projectCwd);
if (harness === 'grok') return normalizeGrokEvent(event, projectCwd);
if (harness !== 'cursor') return event;
const cwd = event.cwd
@@ -1959,7 +2018,15 @@ export async function runHook({ stdinJson, env = {}, cwd = process.cwd(), now =
// findings stop being remembered and a reintroduced one reads as fresh.
// Only the immediate tier is remembered: a deferred finding the per-edit
// pass never reported must still read as fresh to the Stop deep pass.
rememberFindings(cache, sessionId, filePath, immediate);
//
// Grok ignores PostToolUse stdout, so Stop is the user-visible pass.
// Remembering here would dedupe those findings out of Stop. Touch the
// file so Stop has it, and leave the finding list empty.
if (harness === 'grok') {
touchFile(cache, sessionId, filePath);
} else {
rememberFindings(cache, sessionId, filePath, immediate);
}
cacheDirty = true;
if (fresh.length > 0) {
@@ -2163,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2191,22 +2261,36 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
return result({ skipped: 'stdin-empty', durationMs: Date.now() - started });
}
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Only Claude Code sends this field; other
// harnesses omit it, so the strict `=== true` is a no-op for them. This
// guard makes the loop impossible regardless of the finding cache key's
// line-number sensitivity (out of scope here; see findingCacheKey).
const harness = resolveHarness(env, event);
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
const harness = resolveHarness(env, event);
audit.harness = harness;
// Grok fires Stop twice: `end_turn` (the gate that can inject
// additionalContext) then an observe-only `shutdown`. A second deep
// pass would re-emit the same findings. Claude omits `reason`; only
// skip when Grok named a reason that is not end_turn.
if (harness === 'grok' && typeof event.reason === 'string' && event.reason !== 'end_turn') {
return result({ skipped: 'stop-reason', reason: event.reason, durationMs: Date.now() - started });
}
// A Stop event carries no file, so the session cwd is the project.
// Umbrella-dir launches keyed their per-edit cache to the edited file's
@@ -2241,6 +2325,7 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
const freshGroups = [];
let scanned = 0;
let cacheDirty = false;
for (const filePath of touched) {
if (scanned >= STOP_MAX_FILES) break;
if (hasPathTraversal(filePath) || SENSITIVE_PATH.test(filePath)) continue;
@@ -2261,29 +2346,39 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
try { content = fs.readFileSync(filePath, 'utf-8'); } catch { continue; }
let findings;
let detectorThrew = false;
const useHtmlEngine = configuredExt
? configuredExt.engine === 'html'
: (ext === '.html' || ext === '.htm');
if (useHtmlEngine && typeof det.detectHtml === 'function') {
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectHtml(filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
} else {
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; }
try { findings = await det.detectText(content, filePath, scanOptions); } catch { findings = []; detectorThrew = true; }
}
// A detector failure tells us nothing about the file. Leave whatever
// was remembered alone rather than recording an empty scan as truth.
if (detectorThrew) continue;
// Full rule set: no tier split here. Config/inline ignores still apply,
// and the session dedupe drops everything the per-edit pass (or an
// earlier Stop pass) already surfaced.
const filtered = filterFindings(findings || [], content, ext, config);
const fresh = dedupeAgainstCache(filtered, cache, sessionId, filePath);
// Sync to the live scan, including empty. Remembering only `fresh`
// (or skipping the write on a clean Stop) left stale keys in place, so
// a finding that was fixed and later reintroduced never fired again.
rememberFindings(cache, sessionId, filePath, filtered);
cacheDirty = true;
if (fresh.length > 0) {
rememberFindings(cache, sessionId, filePath, fresh);
freshGroups.push({ filePath, findings: fresh });
}
}
audit.scannedFiles = scanned;
if (freshGroups.length === 0) {
if (cacheDirty) persistCache(projectCwd, cache);
return result({ emitted: false, skipped: 'stop-clean', durationMs: Date.now() - started });
}
@@ -2300,8 +2395,8 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
);
commitFooterShown(cache, sessionId, text);
// Fresh findings earn the cache write so the next Stop fire is silent
// unless new issues appear; the notice flags ride along.
// Persist the live finding set so the next Stop fire is silent unless
// new issues appear; the notice flags ride along.
persistCache(projectCwd, cache);
return {
exitCode: 0,
@@ -2337,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+10 -9
View File
@@ -2,15 +2,17 @@
/**
* Impeccable design hook PostToolUse + Stop entry point.
*
* Reads the Claude Code / Codex / Cursor hook event from stdin and routes by
* `hook_event_name`:
* Reads the Claude Code / Codex / Cursor / Grok Build hook event from stdin
* and routes by Stop vs everything else. Claude uses `hook_event_name:
* "Stop"`; Grok uses `hookEventName: "stop"`.
*
* - PostToolUse: runs the immediate-tier detector rules against the touched
* file and emits a system reminder via
* `hookSpecificOutput.additionalContext` when findings exist.
* `hookSpecificOutput.additionalContext` when findings exist. Grok
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
@@ -19,7 +21,7 @@
* subprocess. This file is the thin stdin/stdout adapter.
*/
import { runHook, runStopHook, writeAuditLog } from './hook-lib.mjs';
import { runHook, runStopHook, writeAuditLog, isStopEvent } from './hook-lib.mjs';
async function readStdin() {
if (process.stdin.isTTY) return '';
@@ -28,10 +30,9 @@ async function readStdin() {
return Buffer.concat(chunks).toString('utf-8');
}
function isStopEvent(stdinJson) {
function stdinIsStop(stdinJson) {
try {
const event = JSON.parse(stdinJson);
return event && typeof event === 'object' && event.hook_event_name === 'Stop';
return isStopEvent(JSON.parse(stdinJson));
} catch {
// Malformed stdin falls through to runHook, which audits the skip.
return false;
@@ -48,7 +49,7 @@ async function main() {
let stdinJson = '';
try { stdinJson = await readStdin(); } catch { /* fall through */ }
const run = isStopEvent(stdinJson) ? runStopHook : runHook;
const run = stdinIsStop(stdinJson) ? runStopHook : runHook;
const result = await run({
stdinJson,
env: inheritedEnv,
+23 -3
View File
@@ -1013,6 +1013,27 @@ async function fetchLatestSkillVersion() {
}
}
// Destroy fetch's global undici dispatcher before process.exit(): a live
// keep-alive socket trips a libuv assertion on Windows/Node 24 after a
// successful boot (nodejs/node#56645, issue #573).
async function destroyFetchDispatcher() {
const dispatcher = globalThis[Symbol.for('undici.globalDispatcher.1')];
if (dispatcher && typeof dispatcher.destroy === 'function') {
try { await dispatcher.destroy(); } catch { /* exit regardless */ }
}
}
// Drain the boot payload before process.exit(): a live pipe that has not
// flushed yet is truncated when Node tears down (issue #573 review). Then
// close fetch so Windows teardown does not abort on the keep-alive socket.
async function finishCli(output) {
await new Promise((resolve) => {
process.stdout.write(output, () => resolve());
});
await destroyFetchDispatcher();
process.exit(0);
}
// Two instructions used to sit in one directive: ask, and "if they agree, run
// it". Nothing gated the second on an answer, and the same sentence said to
// continue without waiting, so a run that could never establish agreement was
@@ -1159,8 +1180,7 @@ async function cli() {
appendImageToolsDirective(parts);
appendStalenessDirective(parts, ctx, cliOptions);
if (updateDirective) parts.push(updateDirective);
process.stdout.write(parts.join('\n\n---\n\n') + '\n');
process.exit(0);
await finishCli(parts.join('\n\n---\n\n') + '\n');
}
const parts = [`# PRODUCT.md\n\n${ctx.product.trim()}`];
if (ctx.hasDesign) {
@@ -1206,7 +1226,7 @@ async function cli() {
}
}
if (updateDirective) parts.push(updateDirective);
process.stdout.write(parts.join('\n\n---\n\n') + '\n');
await finishCli(parts.join('\n\n---\n\n') + '\n');
}
function parseCliOptions(args) {
+36 -14
View File
@@ -1252,7 +1252,8 @@ export function resolveHarness(env = {}, event = null) {
if (explicit === 'cursor') return 'cursor';
if (explicit === 'github') return 'github';
if (explicit === 'grok') return 'grok';
if (explicit === 'claude' || explicit === 'codex') return 'claude';
if (explicit === 'claude') return 'claude';
if (explicit === 'codex') return 'codex';
// Grok Build sends camelCase `toolName`/`toolInput`/`hookEventName` and no
// snake_case pair. GitHub Copilot sends camelCase `toolName`/`toolArgs`.
// Check Grok first: the old GitHub heuristic (`toolName` and no
@@ -1265,6 +1266,11 @@ export function resolveHarness(env = {}, event = null) {
return 'github';
}
if (typeof event?.conversation_id === 'string' && event.conversation_id) return 'cursor';
// Codex turn-scoped events carry `turn_id`. Claude Code does not. Detecting
// it here means an already-installed Codex hook emits the Codex Stop
// contract without rewriting the hook command to set IMPECCABLE_HOOK_HARNESS.
// https://developers.openai.com/codex/hooks#stop
if (typeof event?.turn_id === 'string' && event.turn_id) return 'codex';
return 'claude';
}
@@ -2224,8 +2230,11 @@ export const STOP_MAX_FILES = 20;
* { exitCode, stdout, audit, emission? }
*
* Never throws; exits silent (and fast) when the session touched no UI
* files. Output uses the Stop hookSpecificOutput channel: additionalContext
* is delivered to the model and the conversation continues so it can act.
* files. Output goes out on the harness's Stop continuation channel: Claude
* Code and Grok Build read hookSpecificOutput.additionalContext, Codex takes
* a decision: "block" whose reason becomes the continuation prompt. Either
* way the findings reach the model and the conversation continues so it
* can act.
*/
export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), now = Date.now, detector } = {}) {
const audit = { ts: new Date(now()).toISOString(), event: 'Stop' };
@@ -2256,17 +2265,21 @@ export async function runStopHook({ stdinJson, env = {}, cwd = process.cwd(), no
audit.harness = harness;
event = normalizeHookEvent(event, cwd, harness);
// Claude Code's Stop-hook contract: `stop_hook_active` is true when this
// hook is being re-invoked only because a prior invocation kept the turn
// alive (here, via hookSpecificOutput.additionalContext). Re-scanning and
// re-blocking now would loop until Claude Code's consecutive-block cap
// force-ends the turn (issue #400). The prior fire already surfaced the
// findings; whether to act on them is the agent's call. Exit fast with no
// output before any scan. Claude sends `stop_hook_active`; Grok sends
// `stopHookActive`, copied onto the snake_case field above. The strict
// `=== true` is a no-op when the field is absent. This guard makes the
// loop impossible regardless of the finding cache key's line-number
// sensitivity (out of scope here; see findingCacheKey).
// Stop-hook re-entry guard: `stop_hook_active` is true when this hook is
// being re-invoked only because a prior invocation kept the turn alive
// (Claude Code via hookSpecificOutput.additionalContext, Codex via a
// decision: "block" continuation). Re-scanning and re-blocking now could
// loop (issue #400). The prior fire already surfaced the findings;
// whether to act on them is the agent's call. Exit fast with no output
// before any scan. Claude Code and Codex both send this field: Codex
// mirrors the Claude contract (StopCommandInput in
// codex-rs/hooks/src/schema.rs) and latches it true for the rest of the
// turn once a block is honored (codex-rs/core/src/session/turn.rs). Grok
// sends `stopHookActive`, copied onto the snake_case field above. Cursor
// and GitHub Copilot omit the field, so the strict `=== true` is a no-op
// for them. The guard makes the loop impossible regardless of the finding
// cache key's line-number sensitivity (out of scope here; see
// findingCacheKey).
if (event.stop_hook_active === true) {
return result({ skipped: 'stop-hook-active', durationMs: Date.now() - started });
}
@@ -2419,6 +2432,15 @@ export function payload(text, eventName = 'PostToolUse', harness = 'claude') {
if (harness === 'github') {
return JSON.stringify({ additionalContext: text });
}
// Codex shares Claude Code's PostToolUse additional-context shape, but its
// Stop schema rejects unknown fields. Findings that should continue the
// turn must be a top-level blocking decision.
// https://developers.openai.com/codex/hooks#stop (schema of record:
// codex-rs/hooks/src/schema.rs, StopCommandOutputWire)
if (harness === 'codex' && eventName === 'Stop') {
if (!String(text ?? '').trim()) return '';
return JSON.stringify({ decision: 'block', reason: text });
}
return JSON.stringify({
hookSpecificOutput: { hookEventName: eventName, additionalContext: text },
});
+1 -1
View File
@@ -12,7 +12,7 @@
* discards that stdout; the scan still warms the session cache for Stop.
* - Stop: runs the FULL detector rule set over every UI file touched this
* session (the deep pass), deduped against what the per-edit pass already
* surfaced, and emits once via the Stop additionalContext channel.
* surfaced, and emits once via the harness-specific continuation channel.
*
* Contract: never break a turn. Always exit 0. Clean files emit a small ack
* unless quiet mode is enabled; a clean Stop pass is silent.
+56
View File
@@ -1060,6 +1060,27 @@ describe('context.mjs CLI', () => {
assert.match(res.stdout, /detect\.mjs --json <changed targets>/);
});
it('drains stdout before exit when the parent pipe is paused', async () => {
const MARKER = 'END_MARKER_573';
write('PRODUCT.md', `# Acme\n\n${'x'.repeat(256 * 1024)}\n\n${MARKER}\n`);
const child = spawn(process.execPath, [SCRIPT_PATH], {
cwd: scratch,
env: { ...process.env, IMPECCABLE_NO_UPDATE_CHECK: '1', IMPECCABLE_NO_STALENESS_CHECK: '1' },
});
let stdout = '';
child.stdout.on('data', (chunk) => { stdout += chunk; });
child.stdout.pause();
const resume = setTimeout(() => child.stdout.resume(), 100);
const status = await new Promise((resolve, reject) => {
child.on('error', reject);
child.on('close', resolve);
});
clearTimeout(resume);
assert.equal(status, 0);
assert.match(stdout, /END_MARKER_573/);
assert.match(stdout, /RESOLVED_CONTEXT:/);
});
// The build-path preference rides the unified config beside hook and
// detector settings. The local file wins because whether a machine can
// generate images is a property of that machine, not of the committed
@@ -1588,4 +1609,39 @@ describe('context.mjs update check', () => {
assert.equal(typeof cache.lastCheck, 'number'); // stamped so we don't re-poll every boot
assert.equal(cache.latestVersion, undefined); // nothing learned
});
// Targeted live-fetch boot: the Windows abort in issue #573 fired after
// stdout was already complete, so the contract is exit 0 with the full
// context still on stdout.
it('exits 0 after a targeted live-fetch boot writes full context', async () => {
const { srv, host } = await startStub({ skills: '2.0.0' });
try {
const { skillScript, project, env } = setup({}, { host });
fs.writeFileSync(
path.join(project, 'package.json'),
JSON.stringify({ private: true, workspaces: ['packages/*'] }),
);
const jervPi = path.join(project, 'packages', 'jerv-pi');
fs.mkdirSync(jervPi, { recursive: true });
fs.writeFileSync(path.join(jervPi, 'PRODUCT.md'), '# Jerv Pi product\n');
const result = await new Promise((resolveRun, rejectRun) => {
const child = spawn(process.execPath, [skillScript, '--target', 'packages/jerv-pi'], {
cwd: project,
env,
});
let stdout = '';
child.stdout.on('data', (chunk) => { stdout += chunk; });
child.on('error', rejectRun);
child.on('close', (status) => resolveRun({ status, stdout }));
});
assert.equal(result.status, 0);
assert.match(result.stdout, /RESOLVED_CONTEXT:/);
assert.match(result.stdout, /# Jerv Pi product/);
assert.match(result.stdout, /UPDATE_AVAILABLE/);
} finally {
srv.close();
}
});
});
+77 -1
View File
@@ -1366,12 +1366,33 @@ describe('writeAuditLog()', () => {
});
describe('payload()', () => {
it('produces hookSpecificOutput for Claude/Codex', () => {
it('produces hookSpecificOutput for Claude', () => {
const obj = JSON.parse(payload('hello'));
assert.equal(obj.hookSpecificOutput.hookEventName, 'PostToolUse');
assert.equal(obj.hookSpecificOutput.additionalContext, 'hello');
});
it('keeps Codex PostToolUse on the Claude-compatible context channel', () => {
const obj = JSON.parse(payload('hello', 'PostToolUse', 'codex'));
assert.equal(obj.hookSpecificOutput.hookEventName, 'PostToolUse');
assert.equal(obj.hookSpecificOutput.additionalContext, 'hello');
});
it('produces a blocking decision for Codex Stop', () => {
const obj = JSON.parse(payload('hello', 'Stop', 'codex'));
assert.deepEqual(obj, { decision: 'block', reason: 'hello' });
});
it('emits nothing for a Codex Stop with no findings text', () => {
assert.equal(payload('', 'Stop', 'codex'), '');
});
it('keeps Claude Stop on the additional-context channel', () => {
const obj = JSON.parse(payload('hello', 'Stop', 'claude'));
assert.equal(obj.hookSpecificOutput.hookEventName, 'Stop');
assert.equal(obj.hookSpecificOutput.additionalContext, 'hello');
});
it('produces additional_context for Cursor', () => {
const obj = JSON.parse(payload('hello', 'PostToolUse', 'cursor'));
assert.equal(obj.additional_context, 'hello');
@@ -2739,10 +2760,20 @@ describe('resolveTargetFiles()', () => {
describe('resolveHarness() / normalizeHookEvent()', () => {
it('routes explicit env and Cursor conversation_id to cursor harness', () => {
assert.equal(resolveHarness({ IMPECCABLE_HOOK_HARNESS: 'cursor' }), 'cursor');
assert.equal(resolveHarness({ IMPECCABLE_HOOK_HARNESS: 'codex' }), 'codex');
assert.equal(resolveHarness({}, { conversation_id: 'c1' }), 'cursor');
assert.equal(resolveHarness({}, { turn_id: 'turn-1' }), 'codex');
assert.equal(resolveHarness({}), 'claude');
});
it('prefers explicit harness and Cursor detection over the Codex turn_id', () => {
assert.equal(resolveHarness({ IMPECCABLE_HOOK_HARNESS: 'claude' }, { turn_id: 'turn-1' }), 'claude');
assert.equal(resolveHarness({ IMPECCABLE_HOOK_HARNESS: 'grok' }, { turn_id: 'turn-1' }), 'grok');
assert.equal(resolveHarness({}, { conversation_id: 'c1', turn_id: 'turn-1' }), 'cursor');
assert.equal(resolveHarness({}, { turn_id: '' }), 'claude');
assert.equal(resolveHarness({}, { turn_id: 42 }), 'claude');
});
it('maps Cursor postToolUse Write path into file_path + cwd', () => {
const normalized = normalizeHookEvent({
conversation_id: 'c1',
@@ -3886,6 +3917,51 @@ describe('runStopHook()', () => {
assert.equal(stop.emission.kind, 'stop-deep-pass');
});
it('emits Codex Stop findings as a blocking decision', async () => {
const sid = 'stop-codex';
write('package.json', '{}');
const file = write('src/Card.tsx', 'noop');
const det = fakeDetector([finding('marketing-buzzword', 3)]);
const editEventCodex = { ...editEvent(file, sid), turn_id: 'turn-1' };
const stopEventCodex = { ...stopEvent(sid), turn_id: 'turn-1' };
const edit = await runHook({ stdinJson: JSON.stringify(editEventCodex), env: {}, cwd, detector: det });
assert.equal(edit.audit.harness, 'codex');
assert.equal(edit.audit.deferred, 1);
const editOut = JSON.parse(edit.stdout);
assert.ok(editOut.hookSpecificOutput, 'Codex per-edit output stays on the PostToolUse context channel');
assert.equal(editOut.decision, undefined);
const stop = await runStopHook({ stdinJson: JSON.stringify(stopEventCodex), env: {}, cwd, detector: det });
assert.equal(stop.exitCode, 0);
assert.equal(stop.audit.harness, 'codex');
assert.equal(stop.audit.emitted, true, JSON.stringify(stop.audit));
const out = JSON.parse(stop.stdout);
assert.equal(out.decision, 'block');
assert.match(out.reason, /marketing-buzzword/);
assert.ok(out.reason.trim().length > 0, 'Codex ignores a block whose reason trims empty');
assert.equal(out.hookSpecificOutput, undefined);
});
it('skips the Codex Stop re-fire after a block instead of blocking again', async () => {
const sid = 'stop-codex-refire';
write('package.json', '{}');
const file = write('src/Card.tsx', 'noop');
const det = fakeDetector([finding('marketing-buzzword', 3)]);
await runHook({
stdinJson: JSON.stringify({ ...editEvent(file, sid), turn_id: 'turn-1' }),
env: {},
cwd,
detector: det,
});
const refire = { ...stopEvent(sid), turn_id: 'turn-1', stop_hook_active: true };
const stop = await runStopHook({ stdinJson: JSON.stringify(refire), env: {}, cwd, detector: det });
assert.equal(stop.exitCode, 0);
assert.equal(stop.stdout, '');
assert.equal(stop.audit.skipped, 'stop-hook-active');
});
it('keeps a policy footer when the grouped Stop render is clamped to the minimum budget', async () => {
const sid = 'stop-clamp';
fs.mkdirSync(path.join(cwd, '.impeccable'), { recursive: true });
+13 -4
View File
@@ -108,6 +108,12 @@ function createFakeUniversalBundle(root, providers = ['.claude', '.agents', '.cu
hooks: { PostToolUse: [{ matcher: 'apply_patch', hooks: [{ type: 'command', command: 'node ".codex/skills/impeccable/scripts/hook.mjs"' }] }] },
}, null, 2));
}
if (providers.includes('.grok')) {
mkdirSync(join(bundleRoot, '.grok', 'hooks'), { recursive: true });
writeFileSync(join(bundleRoot, '.grok', 'hooks', 'impeccable.json'), JSON.stringify({
hooks: { PostToolUse: [{ matcher: 'Edit|Write|MultiEdit', hooks: [{ type: 'command', command: 'node ".grok/skills/impeccable/scripts/hook.mjs"' }] }] },
}, null, 2));
}
// Native subagent definitions, mirroring the build's provider agents output.
if (providers.includes('.github')) {
mkdirSync(join(bundleRoot, '.github', 'agents'), { recursive: true });
@@ -1125,21 +1131,24 @@ describe('skills install/update: local universal bundle e2e', () => {
const tmp = mkdtempSync(join(tmpdir(), 'imp-test-scope-user-hooks-'));
const home = mkdtempSync(join(tmpdir(), 'imp-home-scope-user-hooks-'));
execSync('git init', { cwd: tmp });
const bundleRoot = createFakeUniversalBundle(tmp, ['.claude', '.agents', '.cursor']);
const bundleRoot = createFakeUniversalBundle(tmp, ['.claude', '.agents', '.cursor', '.grok']);
const output = run('skills install -y --providers=claude,codex,cursor --scope=global', {
const output = run('skills install -y --providers=claude,codex,cursor,grok --scope=global', {
cwd: tmp,
env: { ...process.env, HOME: home, IMPECCABLE_BUNDLE_PATH: bundleRoot },
});
expect(output).toContain('Installed impeccable into: .claude, .agents, .cursor (global)');
for (const provider of ['.claude', '.agents', '.cursor']) {
expect(output).toContain('Installed impeccable into: .claude, .agents, .cursor, .grok (global)');
for (const provider of ['.claude', '.agents', '.cursor', '.grok']) {
expect(existsSync(join(home, provider, 'skills', 'impeccable', 'SKILL.md'))).toBe(true);
expect(existsSync(join(tmp, provider, 'skills', 'impeccable', 'SKILL.md'))).toBe(false);
}
expect(readFileSync(join(tmp, '.claude', 'settings.local.json'), 'utf8')).toContain(join(home, '.claude', 'skills', 'impeccable', 'scripts', 'hook.mjs'));
expect(readFileSync(join(tmp, '.codex', 'hooks.json'), 'utf8')).toContain(join(home, '.agents', 'skills', 'impeccable', 'scripts', 'hook.mjs'));
expect(readFileSync(join(tmp, '.cursor', 'hooks.json'), 'utf8')).toContain(join(home, '.cursor', 'skills', 'impeccable', 'scripts', 'hook-before-edit.mjs'));
const grokHooks = readFileSync(join(tmp, '.grok', 'hooks', 'impeccable.json'), 'utf8');
expect(grokHooks).toContain(join(home, '.grok', 'skills', 'impeccable', 'scripts', 'hook.mjs'));
expect(grokHooks).not.toContain('".grok/skills/impeccable/scripts/hook.mjs"');
rmSync(tmp, { recursive: true, force: true });
rmSync(home, { recursive: true, force: true });