mirror of
https://github.com/pbakaus/impeccable.git
synced 2026-09-11 21:57:14 +03:00
main
40
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  Paul BakausandGitHub
|
05ef444b7d |
Use GPT Image 2.5 Flare and native transparent assets (#786)
* Use GPT Image 2.5 Flare for image generation Add explicit model selection and cover generation/edit request provenance. Compare 24 comp samples with Image 2 and Sunburst, documenting efficiency gains and the photo-preservation tradeoff. Prepared with AI assistance under maintainer instructions. * Validate native image transparency for asset production Add repeatable cutout, glass, and reference-plate experiments with alpha measurements and compositing evidence. Record fidelity limits and the remaining asset-producer integration work. Prepared with AI assistance under the maintainer instructions. * Add native transparent PNG asset generation Wire background options through generation and edits, preserve alpha, and update asset production to use supported commands with visual cutout checks. Prepared with AI assistance under the maintainer's instructions. * Fix native alpha review findings Validate PNG paths for all background modes, score sparse and partial alpha over the sampled ground, and correct launcher and plate workflow guidance. Prepared with AI assistance under the maintainer's instructions. * Fix launcher paths in comp workflow commands Quote existing launcher filenames for the target shell while retaining multiword command prefixes. Cover relative and absolute paths and actual shell argument parsing. Prepared with AI assistance under the maintainer's instructions. * Remove evaluation artifacts and docs from image PR Keep the PR focused on runtime behavior, source skill instructions, and regression coverage. Preserve evaluation material locally. Prepared with AI assistance under the maintainer's instructions. |
||
|
Paul BakausandGitHub
|
6496f49a1e |
Fix skill workflow regression coverage (#783)
Clarify launcher fallback and completed documentation handoffs; separate bounded protocol checkpoints from opt-in browser-backed completion diagnostics. Correct fixture containment, target syntax, and artifact assertions. AI assistance: Codex, under maintainer direction. |
||
|
e2ff625b63 |
The Rust engine: one binary replaces every script and the JS detector, fully open (#714)
* Add oracle harness: verb goldens and function-level vectors
Records stdout/stderr/exit/files for every impeccable verb over a fixed
corpus and replays them against an alternate implementation. Adds a loader
hook that captures per-function call vectors from the pure engine modules.
Prepared with AI assistance (Claude Code).
* Oracle: hook, hook-before-edit, hook-admin cases and goldens
Prepared with AI assistance (Claude Code).
* Add docs/CLI-CONTRACT.md: observable behavior of every impeccable verb
Prepared with AI assistance (Claude Code).
* Oracle: context/doctor/pin/surface-brief/critique/palette/embed/signals/csp/seed/genimg/question cases and goldens
Prepared with AI assistance (Claude Code).
* Oracle: live-mode cases and goldens (roots, inject, wrap, insert, accept, session, manual edits, daemon)
Prepared with AI assistance (Claude Code).
* Oracle: mask the binary path before HOME; export launcher env to the binary
Prepared with AI assistance (Claude Code).
* detect: set process.exitCode instead of exiting after the final write
process.exit() right after a large piped stdout write truncated JSON output
at the pipe buffer boundary; found by the oracle harness. Re-record the six
directory-scan goldens that had captured the truncation.
Prepared with AI assistance (Claude Code).
* Oracle: normalize the hook-admin command in both runtimes' forms and audit chars
Prepared with AI assistance (Claude Code).
* Skill text: invoke the impeccable launcher instead of node scripts
Every `node {{scripts_path}}/<name>.mjs` becomes `{{scripts_path}}/impeccable <verb>`
(context-signals -> signals, hook-admin -> hooks). Setup step 1 drops Node, points
Windows shells without sh at impeccable.cmd, and says the launcher runs a
self-contained binary. allowed-tools follows.
Prepared with AI assistance (Claude Code).
* Scripts dir: replace the Node scripts with the impeccable launcher
skill/scripts keeps command-metadata.json and the page JS; every .mjs entry
point, lib/, and live/ are gone (the binary owns those verbs). Adds the POSIX
launcher, impeccable.cmd, VERSION (copied from the new root ENGINE_VERSION),
scripts/fetch-engine.mjs (bun run fetch:engine) to pull the pinned binary
into skill/scripts/bin/<os>-<arch>/, and gitignores that bin dir.
Prepared with AI assistance (Claude Code).
* Build: ship the launcher instead of bundling the JS engine
readSourceFiles no longer copies cli/engine into the skill; the scripts
payload is the launcher (executable bit preserved through dist, plugin/, and
universal.zip), impeccable.cmd, VERSION (synced from ENGINE_VERSION on every
build), the page JS, and command-metadata.json. Hook manifests call
`<scripts>/impeccable hook` behind an existence guard (Codex adds a
commandWindows sibling calling impeccable.cmd; Cursor runs hook-before-edit;
GitHub keeps the git rev-parse form; Grok mirrors Claude); the Node probe and
systemMessage notice are gone. build:release fetches the pinned engine for
every target (lenient) and stages bin/<os-arch>/ into the dist skill copies
after root harness dirs and plugin/ were synced, so git-delivered trees stay
launcher-only. The detection-rule count check reads the vendored
extension/detector/antipatterns.json and is skipped when absent.
build:browser is a stub; the codex prefix rewrite leaves
`{{scripts_path}}/impeccable` alone.
Prepared with AI assistance (Claude Code).
* CLI: turn the impeccable npm package into a platform-binary shim
cli/engine, cli/lib, and cli/bin/commands are gone; their behavior lives in
the engine binary. cli/bin/cli.js now resolves the binary from IMPECCABLE_BIN,
the @impeccable/cli-<os>-<arch> optional dependency (templates under
cli/platform-packages/, published by the engine release), the
~/.impeccable/bin/<version>/ cache, or a checksum-verified download, and
execs it. package.json drops the engine dependencies and the library
exports; puppeteer moves to devDependencies for the icon scripts.
README.npm.md describes the shim.
Prepared with AI assistance (Claude Code).
* Tests: gate behavior on the oracle and the engine binary
Unit tests of the deleted Node scripts and the JS detector are removed;
their behavior is pinned by tests/oracle goldens (frozen JS behavior plus
reviewed deltas) and the engine's own tests. tests/oracle.test.mjs replays
the corpus against the binary (IMPECCABLE_BIN or skill/scripts/bin/<target>/,
via tests/lib/engine-bin.mjs) and skips cleanly without one; the framework
fixture sweep drives live-inject, live-wrap, and detect-csp through the
binary the same way. record.mjs learns --bin. The function-level vectors
under tests/oracle/vectors/calls are committed as the frozen snapshot they
can no longer be regenerated from. Suites: core trimmed to build and
transformer tests, oracle added to the default run, detector/live reduced to
packaging and reference checks, the live-e2e helper tests move to the opt-in
live-e2e lane pending its retarget, cli-remote-e2e is an empty placeholder.
Prepared with AI assistance (Claude Code).
* Docs: describe the launcher, the engine pin, and the oracle gate
CLAUDE.md gains an Engine binary section (launcher lookup order, ENGINE_VERSION,
untracked binaries, how tests get one, the oracle as behavior gate, what stays
JavaScript) and drops the Node-script and JS-detector descriptions; the CLI
and detection-rule sections point at the shim and the engine repo. README.md
states the skill needs no runtime and lists the launcher-based hook commands;
AGENTS.md follows. CLI-CONTRACT.md's intro notes the scripts it quotes are
the recorded source, not the tree.
Prepared with AI assistance (Claude Code).
* Tests: tighten the hook command guard assertion
Prepared with AI assistance (Claude Code).
* Oracle: re-golden 46 cases for the engine's own command names; record them in DELTAS.md
Prepared with AI assistance (Claude Code).
* Build: ship launcher-only release zips by default
IMPECCABLE_BUNDLE_ENGINE=1 opts in to staging the engine binaries into the
dist skill copies. Bundling every target into every provider copy put
dist/universal.zip near 340 MB, past the 25 MB Cloudflare Pages file cap
that impeccable install downloads through.
Prepared with AI assistance (Claude Code).
* Tests: drive the live-e2e orchestrator through the engine binary
The session, fake-agent loop, steer test, and manual-edit probe spawn
<binary> <verb> (live-server, live, live-inject, live-wrap, live-insert,
live-accept, live-poll, live-complete) resolved by tests/lib/engine-bin.mjs
instead of node skill/scripts/live-*.mjs; the completion typing the agent
imported from the deleted live/completion.mjs is a small local helper. The
live-e2e helper unit tests move back into the default live suite (the steer
loop skips without a binary).
Prepared with AI assistance (Claude Code).
* Tests: run new-work-e2e through the engine's serve-question and generate-image verbs
Prepared with AI assistance (Claude Code).
* Tests: point the skill-behavior harness at the launcher and engine binary
The bash tool exports IMPECCABLE_BIN so the staged skill's launcher runs
without a download; scenarios assert on 'impeccable context' instead of
context.mjs and skip without a binary.
Prepared with AI assistance (Claude Code).
* Tests: note what plugin-e2e validates before and after the generated-output sync
Prepared with AI assistance (Claude Code).
* Oracle: record the engine's 'wasm-unsafe-eval' CSP meta patch as a reviewed delta
Prepared with AI assistance (Claude Code).
* Rebase reconciliation: fold main's post-freeze work into the swapped tree
The rebase onto origin/main brought changes whose JS engine halves left the
tree with the swap. This commit reconciles what survives:
- Suite map: register main's comp-fidelity unit tests (build-phase,
comp-diff, font-match, hero-checks) in the core suite and
live-browser-ignores in the live suite.
- Payload guard: the skill scripts payload now allowlists the comp-fidelity
build pipeline (comp-spec/comp-diff/build-phase/font-match and their libs),
the one Node toolchain that has not moved into the engine.
- Drop skill/scripts/live/project-ignores.mjs, lib/live-path-globs.mjs, and
their test: they import hook-lib/live-inject/impeccable-paths, which the
swap deleted, and their consumer (the JS live server) is the engine now.
- skill text: the comp pipeline's calls to engine verbs (generate-image,
embed-prompt) use the launcher spelling.
- Oracle: re-record 17 detect goldens over the fixture set main changed
(oklch #592, color-mix #578, 1D grid #615, the two comp-fidelity rules)
and record the gap in DELTAS.md; those JS rule changes are not yet ported
to the engine, and the goldens pin its current behavior.
bun run test (oracle included) and bun run build are green on this tree.
AI-assisted change: implemented with Claude Code.
Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WaJv2c4oN8wS7Ttq4XRqyx
* Launcher: engine-probe PATH validation, working .cmd download path; CI: drop stale path, add oracle job
Byte-identical copies of the engine repo's launchers (engine main
af7572c): the retired 3.x npm CLI on PATH or in ~/.impeccable/bin is
rejected by the engine-probe handshake instead of hijacking every verb;
impeccable.cmd's download path is rewritten as straight-line goto flow
(the parenthesized blocks expanded %url%/%cached% at parse time, making
it dead code) with certutil sha256 verification and a windows-arm64 ->
x64 asset fallback; the final error points at the release download
instead of npm i -g (npm still serves the 3.x CLI).
ci.yml: the generated-output check no longer diffs the deleted
cli/engine/detect-antipatterns-browser.js, and a new oracle job fetches
the pinned engine (bun run fetch:engine) and replays tests/oracle/
against it. The job is continue-on-error with a loud warning until the
first engine release exists; flipping it to required is a release-time
toggle, documented in the workflow.
Verified here: sh -n on both launcher copies, bun run build green, full
oracle replay against the rebuilt engine binary green (770 pass, 0
fail), and a launcher behavior test proving a fake 3.x CLI on PATH is
skipped while the download + checksum chain completes against a local
file server.
Prepared with AI assistance (Claude Code).
* Oracle: restore detector goldens to post-fix behavior after the engine ports
The Aug 17-31 detector fixes (oklch parsing, color-mix nested hex, 1D grid
pass, comment stripping, root-relative linked stylesheets, URL userinfo
redaction, inert ignore-value refusal) and the comp-fidelity rules
organic-clip-path / buried-raster are ported to the engine. Re-records the
gap-pinning detect goldens from the fixed binary (glow.html included: its
.photo-opaque-grad column now carries the buried-raster finding it was
written for), replays the frozen checkHtmlPatterns call vectors through the
last JS engine state in history (db1462b9^; args untouched, 14 of 101
results moved), and rewrites the DELTAS gap section into the landed-ports
note. Each re-recorded json fixture golden byte-matches that JS state's
output; oracle: 770 pass, 0 fail.
Prepared with AI assistance (Claude Code).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WaJv2c4oN8wS7Ttq4XRqyx
* Oracle: pin the Aug 17-31 verb fixes ported to the Rust engine
New cases: hook-session-grok-edit-then-stop (Grok Build camelCase envelope,
end_turn/shutdown/stopHookActive Stop handling,
|
||
|
6dd15238c4 |
Close the comp-round gap and make the hero gate teach
concept-seed's choice ping now prints the next mandatory step from the recorded build path (comp-led: build-phase.mjs start --direction <key>), because every run that skipped the comp round did so right after that ping. build-phase gains a comps phase ahead of spec (three sidecar'd comps under .impeccable/mocks/, one approved) and records the approved comp on close. The hero gate lists the worst region crops first with the fix class per verdict, and refuses a third value-only attempt on the same stuck region. Hero instruction is plates first, then the semantic layer. The finish reviewer treats a comp-led build with no closed comps phase as a material finding. AI-assisted (Claude). Co-Authored-By: Claude <noreply@anthropic.com> |
||
|
|
34ef9ac2c4 |
Rewrite the comp-led build around the phase gates; reviewer reads the diff first
new-work.md section 6 becomes the spec / plates / hero / sections / motion / responsive phase list, each closed by build-phase.mjs advance; the reproduction and hero-checkpoint prose that asked the model to compare from memory is gone. visualize.md's inventory, medium gate, and produce sections collapse into the spec and plate mechanism. The finish reviewer takes the state file and comp-diff reports as inputs and starts its fidelity matrix from the measured verdicts. docs/COMP-FIDELITY.md records the design. AI-assisted (Claude). Co-Authored-By: Claude <noreply@anthropic.com> |
||
|
|
5856161014 |
Plate pipeline, asset producer rewrite, and two detector rules for CSS standing in for material
generate-image.mjs --plate produces one raster region of the measured spec from the comp crop, scores it against the crop, and refuses under --min. The asset producer's job becomes producing the spec's plates. Detector gains organic-clip-path (many-vertex polygon / curved path() clips) and buried-raster (raster under a near-opaque wash or at near-zero opacity), wired into both engines with fixtures. AI-assisted (Claude). Co-Authored-By: Claude <noreply@anthropic.com> |
||
 Abdul WahabandCursor
|
e9c62278c1 |
Make the code-led GROUND fallback deterministic, compare like for like
The quality bar leaves the color-authority chain (it arrives as card image paths and never governs composition). With no comp, a color OWN-WORLD names is the target; when it names none, the review states there is no GROUND authority instead of inventing a target. The build side of the numeric comparison now samples the same way each record was taken: patch average against patch average, gradient ends against gradient ends. AI-assisted change (Cursor), prepared under maintainer direction. Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|
Abdul WahabandCursor
|
79c648a9ab |
Resolve bot review: code-led GROUND authority, sampling rules, tolerance
GROUND no longer lapses silently on code-led builds: with no comp to sample, the authority is the colors OWN-WORLD and the quality bar name, and no invented target beyond them. Non-uniform fields get sampling rules (interior pixel, patch average for texture, both ends of a gradient, never an edge), and the numeric comparison gets tolerance semantics so render noise never fails a faithful build. The hunt hint names the dark-ground prior beside the light one. AI-assisted change (Cursor), prepared under maintainer direction. Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|
Abdul WahabandCursor
|
9213bf1511 |
Generalize color sampling beyond the cream-ground case
Accents join the sampled record alongside ground and dominant fields, every recorded color (not only the ground) is compared by number during the build, and the light-ground-only rationale clauses become value-neutral so dark and saturated comps get the same protection. Rule anchor renamed to skill-color-by-number to match its scope. AI-assisted change (Cursor), prepared under maintainer direction. Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|
Abdul WahabandCursor
|
5b7c9e93cb |
Fix uncaught ground-color drift on comp-led builds
Sample the approved comp's ground and dominant-field hexes into the brief (visualize.md), judge the built page's ground by number against that record including the net value under textures (new-work.md), and make GROUND a mandatory fidelity-matrix row beside TYPE and MATERIAL (finish reviewer). Pre-comp palette chips are retired at approval. AI-assisted change (Cursor), prepared under maintainer direction. Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|
Paul BakausandGitHub
|
49d8cbff16 |
Comp-fidelity review discipline + conciseness pass on core references (#586)
* Comp-fidelity review discipline + conciseness pass on core references Process fixes derived from a real Codex session (Hanasaku landing page) where a build drifted wholesale from the approved comp and still shipped under a reviewer pass: - finish reviewer: new Evidence check (check 0) with a fourth disposition, recapture, for malformed screenshots; a review on invalid evidence binds nothing and owes a full re-review, not a verdict pass - finish reviewer: verdict passes exit scoring mode when recaptures fail check 0 or when the packet carries user-supplied screenshots that contradict a prior verdict (those force a fresh full review); a ship earned in a verdict pass covers the scored fixes, not the whole surface - new-work: capture-validity rules (settle entrance motion, capture from document top, comp comparison at comp dimensions, open every file once before sending); user's actual viewport joins the inspected sizes - new-work: hero checkpoint now writes .impeccable/review/hero-repro.png and the reviewer verifies it exists under Persistence - new-work: comp authority is explicit (only the user can downgrade it); handoff reports the verdict at its actual scope; user evidence reopens a full review; documenter re-runs when fixes land after documentation - craft-floor: Refuse entry for geometric masks approximating organic photographic contours (the circular-cutout failure) - editorial conciseness pass over new-work.md, visualize.md, and both agent files: tighter sentences, no dropped rules, all rule markers and mechanical tokens preserved Assisted-by: Claude Code * fix: define the ship disposition in new-work's action paragraph Copilot review finding: the paragraph claimed exactly four disposition words but defined only recapture, rebuild, and fix. Assisted-by: Claude Code * fix: rebuild returns get a full review; recapture return shape in preamble Cursor Bugbot findings: - a return following a rebuild directive is now a fresh full review on both sides of the contract, never a verdict pass, so a wholesale rebuild cannot earn a scoped ship on the directive alone - the turn-ceiling preamble now names the recapture return shape instead of contradicting it with "the five sections" Assisted-by: Claude Code * fix: absent required captures fail the evidence check Greptile finding: a packet with no desktop.png/mobile.png (or missing native device-class captures) routed to the missing-input notice and could still reach ship. A required capture that is absent now fails check 0 exactly like a malformed one and forces recapture; the missing-input allowance in the preamble excludes captures. Assisted-by: Claude Code * fix: user-viewport capture is a required, named input to the review Greptile finding: the evidence gate hard-coded web requirements to desktop.png and mobile.png, so a reported user viewport could join the inspected set and still ship uncaptured. The parent now saves it as user-<width>.png and names every inspected viewport required in the packet; check 0's required set includes every brief-named capture. Assisted-by: Claude Code |
||
|
Paul BakausandClaude Fable 5
|
cb305fdca1 |
Two adversarial reviews later, the discipline says half as much
Two independent skeptic passes over the added prose, one hunting oversteer and example bias, one hunting mode and platform damage. What they killed, and why: - The absolute 'never a medium' rule contradicted the file's own imagery-stance fixity two paragraphs up and stripped legitimate guards (an illustration-committed world, a native app screen warding off stock-photo drift). A medium ban now belongs to the committed imagery stance, never to caution, and the rule appears once per reader context instead of five times corpus-wide. - The quoted incident string and the four-example subject list taught the model the exact framings they existed to prevent. Gone; the abstract rule plus the point-at-the-subject check carry it. - 'A first-time visitor learns what this is, why it matters, and what to do' was Persuade anatomy imposed on all four modes. The guard is now mode-neutral: a quieted region keeps its information and stops performing. - 'Calm is what Operate and Read surfaces are for' contradicted operate.md's density affordance. Deleted; modes stay defined in one place. - The focal-moment count now presupposes nothing: it fires only where the direction names a focal moment, and only on same-scale rivalry, so an even, calm field stops reading as a failure. - The decision-comp clause and the mode bullet no longer restate what they can reference. Net: the prose additions drop from roughly 480 words to under 200, with no quoted strings and no example lists. AI-assisted (Claude Fable 5), prepared for maintainer review. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
e867487d55 |
Ban fabrications, never media: counter the exclusion-list reflex everywhere prompts are authored
The declined moto-forum comp's prompt read 'no gradients, no rounded SaaS cards, no photography, no fake member counts, no badges, no testimonials': the reflex that rightly bans invented claims swallowed the one medium the subject lives in, and that is exactly how a motorcycle forum got comped with no motorcycles. The lektor prompt's 'no AI imagery', written by an image model, is the same fingerprint. One counterweight, phrased once per authoring surface: the comp discipline's subject-presence check (which the decision comps already bind), the asset producer's own prompt rules (a standalone agent that never reads visualize.md), and new-work's author-assets law (the path a code-led build takes without the comp round). Truth binds claims, not demonstrations; a photo of the subject doing its job is a demonstration. AI-assisted (Claude Fable 5), prepared for maintainer review. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
248a4a699a |
Retire the sketch era's wire name: the field is comp, sketch is an alias
The deliverable died in #545; the word survived as the decision-page payload's field name, annotated everywhere it appeared with the same compatibility apology. The page and the skill text ship together and payloads are per-session, so the compatibility burden is one input alias, not a frozen name. serve-question.mjs: the card field, the answer key, the schema docs, the --schema example, the help text, and every internal identifier (compSrc, data-comp, .media.comp-pending, img.comp, comp-note) now say comp; a payload declaring the legacy sketch key still renders and answers identically. new-work.md and the asset producer drop their wire-name parentheticals. The unit suite covers the canonical answer key coming back from a legacy-key payload; the new-work e2e's declined-card stray comp stays declared as sketch, which doubles as alias coverage. AI-assisted (Claude Fable 5), prepared for maintainer review. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandGitHub
|
19786e7a22 |
Native leg for the verify-and-review pipeline (#546)
* Give the verify-and-review pipeline a native leg The build-verify-review loop assumed a browser end to end while the comp side of the system was already platform-aware: new-work.md, visualize.md, and the asset producer all comp a native app portrait at its device viewport, and then the verification steps asked for desktop and mobile browser screenshots of it. Concretely: - new-work.md step 7 ordered detect.mjs on every hookless build with no platform guard. routing.md declares the detector web-only and the design hook skips native projects, so a native build was always hookless and always ordered to run an HTML rule engine over Swift/Kotlin/RN code. The playbook now guards it: web-only, and on native the reviewer's floor check is the named slop gate. - The inspection round and the SKILL.src.md batched-round principle named desktop and mobile as the only viewports. Both now map per platform: web keeps desktop and mobile; native inspects the shipped device classes per OS, captured from the simulator or emulator. - ios.md and android.md carried no verification guidance at all, so nothing told a native run how to produce the screenshots the evidence chain depends on. Each gains a Verifying the build section: simctl / adb capture commands, dark-appearance and type-scale checks, and the simulator-vs-hardware honesty line. - The finish reviewer judged native builds blind: it never runs context.mjs and its packet carried no platform guidance. On native the packet now includes the platform reference path(s) and a no-detector-ran line, and the reviewer's Input Contract says to judge in the platform's conventions. Assisted-by: Claude Code * fix: address PR review bot findings - greptile: carry the capture's device selector through the state-changing verification commands (simctl appearance, adb uimode and font_scale); unqualified forms fail with several targets attached - Copilot: align new-work.md's cross-reference with the actual heading (Verifying the build) - Copilot: give the finish reviewer's Input Contract the native filename example new-work.md establishes (phone.png / tablet.png, suffixed per OS on adaptive) Assisted-by: Claude Code * fix: identify simulators by UDID, not display name greptile follow-up: display names can collide across booted simulators, so the capture and appearance commands now both key on the UDID from simctl list devices booted. Assisted-by: Claude Code |
||
|
|
045865918a |
Held for review: agent placeholder substitution, reviewer recapture contract, base-directory script form (#544)
* Resolve {{scripts_path}} in the agent bodies Codex ships
Three code paths emit an agent body: the degraded fallback reference, the
.toml nested inside the skill for Codex, and the native agent file. Only the
nested .toml skipped placeholder substitution and rule-marker stripping, so
the codex and .agents dists shipped `node {{scripts_path}}/embed-prompt.mjs`
verbatim in the asset producer, and every caller had to substitute the token
itself at load time.
All three now render through renderAgentBody(), and the new regression test
asserts a runnable embed-prompt command on each emitted surface plus a
synthetic agent proving markers and placeholders resolve in the nested .toml.
Prepared by an AI agent (Claude Code) under pbakaus's instruction.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Give the finish reviewer's screenshots one fixed address
The Input Contract asked for "desktop and mobile screenshot paths captured by
the parent" and named none, so each session invented a filename and the
verdict pass went looking for a recapture that was never written there. Two
reviewer passes burned on that in the eval runs.
The parent now captures and recaptures to .impeccable/review/desktop.png and
.impeccable/review/mobile.png, and the reviewer reads those two first,
treating a brief-named path as the fallback for a parent that wrote elsewhere.
Prepared by an AI agent (Claude Code) under pbakaus's instruction.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Lead Setup with the base directory the runtime reports
The rendered claude and codex skills opened with
`node .claude/skills/impeccable/scripts/context.mjs`, a project-relative path
that resolves in this repo and in nothing a user installs: a personal or
plugin install puts the scripts outside the project entirely. The working form
was already in the text, parenthesized, after the one that fails.
Setup now leads with `node <skill-base-dir>/scripts/context.mjs` and says once
that the base directory resolves every scripts-path command in the skill and
its references, leaving the project-relative path as the fallback for runtimes
that report no base directory.
Prepared by an AI agent (Claude Code) under pbakaus's instruction.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Answer the Copilot review: brittle model assertion, missing review dir
Assert that {{model}} resolved rather than that it resolved to "GPT", which
belongs to PROVIDER_PLACEHOLDERS and can change without touching what the test
guards. And have the parent create .impeccable/review/ when the harness does
not, so a fresh project's first capture has somewhere to land.
Prepared by an AI agent (Claude Code) under pbakaus's instruction.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Make the review-screenshot contract directory-based, not web-viewport-named
Two amendments to the recapture contract from review feedback:
1. The canonical location is the directory .impeccable/review/, one file
per captured viewport; desktop.png and mobile.png are the web case,
not the contract. Baking web-viewport names into the reviewer's spec
would have hardened a web assumption into paths that a native
(ios/android/adaptive) build cannot honestly write.
2. Precedence restored to explicit-beats-convention: paths the calling
brief names are authoritative when the files exist; the canonical
directory is where the reviewer looks when the brief names none or a
named path is missing. This avoids stale canonical files from an
earlier run silently winning over fresh explicit paths. The observed
failure (the verdict round inventing a round-stamped filename) stays
fixed: recapture happens over the same files, and invented filenames
are still called out as pointing at nothing.
Assisted-by: Claude Code
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
||
|
|
dbff0880e6 |
Decision page: full-fidelity comps, raise cycler, declined sizing, canon order, full card anatomy (#545)
* Polish the decision page: raise cycler, declined height, canon order, full card anatomy Field feedback from the first real rolls of the verdict-routed hand: - Several raises stacked on the assigned card blew it out of proportion. More than one raise now renders as a compact cycler: one visible, a counter, click or Enter advances. A single raise stays inline. - Declined cards inherited the row's stretch alignment, so a narrow card stood at the tallest contender's height, a strange stilt beside the hand. They now size to their content. - Deck order becomes a gradient of standing: contenders, then the canon, then declined dead last. The canon between full alternates and the demoted row reads as the familiar door rather than the last resort after the rejects. - Root cause of bare-bones challenger and canon cards in the field: the --schema example only gave the assigned card palette, materials, and risk, and models author payloads by imitating the example, so the "same anatomy on every card" instruction lost to it every time. The example now carries full anatomy on every card and the schema note says a card with no palette chips is an authoring gap, not a data gap. AI-assisted change. Co-Authored-By: Claude Code <noreply@anthropic.com> * Decision cards carry full-fidelity comps instead of sketches Field verdict on the sketch contract: the sketches came back too simple to inform the choice, and generation takes the same time at any fidelity, so the deliberately-unfinished frame paid comp cost for sketch quality. The decision card's image is now that direction's north-star comp, produced under visualize.md's comp discipline (structure-led prompt, real name and content, no invented commercial claims), saved under .impeccable/mocks/ with its prompt sidecar. Fairness between cards comes from equal fidelity in each card's own grammar rather than shared unfinishedness. The chosen card's comp is never spent by the choice: on a comp-led build it enters the comp round as compositional option one (visualize.md now generates two variations beside it; a round arriving with no decision comp still renders all three), and on a code-led build it returns at the finish review as the critique reference. Produce order still front-loads a re-roll's spend onto the cards read first. serve-question keeps the sketch field's wire name for payload compatibility; docs, schema paths, shimmer labels, and the answer directive (CHOSEN COMP) speak comp. AI-assisted change. Co-Authored-By: Claude Code <noreply@anthropic.com> * fix: address PR review bot findings on the comp round - Producer still forced sketches (cursor, high): the asset producer's Decision Sketches contract still mandated deliberately unfinished matte sketches, so the parallel path would keep shipping sketch-era images. The section is now Decision Comps: full-fidelity north-star comp, structure-led prompt, equal commitment across siblings, no invented claims, sidecar written. - Mocks collided with the approval check (cursor, high): decision comps now live under .impeccable/mocks/decision/, visualize.md scopes the no-approval finding to comp-round output, new-work.md states the unchosen hand implies no approval, and the code-led finish packet names the chosen decision comp as the critique reference in the approved-comp slot. - Raise cycler announces (greptile, both P1s): a visually hidden aria-live region reads out the newly active raise and its position on advance; initial render stays quiet. - Declined width in the vertical deck (cursor, medium): align-self: flex-start shrank declined cards to content width in the portrait column layout, where the cross axis is horizontal; they stretch there and keep content height in the row layout. AI-assisted change. Co-Authored-By: Claude Code <noreply@anthropic.com> * fix: raise cycler tooltip and label name both input modes Copilot: the tooltip said Click while the control also answers Enter and Space; the title and a new aria-label now say activate/press Enter. AI-assisted change. Co-Authored-By: Claude Code <noreply@anthropic.com> * fix: finish reviewer exempts decision comps from the approval check cursor[bot] follow-through: the reviewer's Persistence check still treated any comps under .impeccable/mocks/ as approval-gated, and the reviewer never reads visualize.md by design, so code-led and spent-hand rounds could draw a false skipped-approval finding. The check now scopes to comp-round comps, exempts .impeccable/mocks/decision/ as the direction round's dealt hand, and defines how a code-led build's decision comp is judged in the approved-comp slot: the critique reference, under the no-approved-comp fidelity rules plus what the image dared that the build did not. AI-assisted change. Co-Authored-By: Claude Code <noreply@anthropic.com> * fix: the critique reference is its own reviewer input, not the approved-comp slot cursor[bot]: passing the code-led decision comp through the approved-comp slot dragged in that slot's obligations (inventory-first reading, the fidelity matrix, Truth's shipped-asset demand for every image-native region), which contradicts code-led's premise. The input contract now names it a separate labeled critique-reference input that nothing binding "the approved comp" touches, and Fidelity defines its treatment where the no-approved-comp rules live: provocation, not spec; no matrix, citations, or asset obligations; its dares enter material_fixes as ordinary fixes. AI-assisted change. Co-Authored-By: Claude Code <noreply@anthropic.com> --------- Co-authored-by: Claude Code <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
827dfeb95e |
The primary action is signature material, not chrome
The Tortuga comp dissolves the Install CTA's edge into the storm's particles; the build shipped a plain rectangle with four decorative dots, and neither the builder nor the reviewer's rebuild findings named it. The inventory now gives the primary action its own row and medium, naming the shrink-to-border-trick failure as the compliance-token version of commitment, and the reviewer's fidelity matrix lists the primary action's treatment among the salient elements, with a physically-worked CTA rendered as a plain rectangle scored contradicted. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
861682eebb |
Give the finish reviewer and documenter craft-floor authority
Codex's own post-mortem of the second hamster-wheel session: it loaded the kicker ban, shipped five kickers anyway, and then the reviewer and documenter 'compounded it by accepting, and even canonizing, the invented label style'. Nothing downstream of the builder ever re-read the floor. The reviewer gains check 6, Floor: read craft-floor.md (now the one skill reference it may read, passed in its inputs) and hold the screenshots against the Refuse list; a banned element is a material fix even when it matches nothing in the comp, because fidelity cannot authorize what the floor refuses. The documenter gains the mirror rule: a floor refusal lands in its not-canonized line as a carried defect, never in DESIGN.md as a rule future surfaces inherit. Also updates the live-reference contract test to match the applier's condensed no-server sentence, which still carries the same guarantee. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
28af30eff0 |
Condense the grown skill files and harden the reviewer's verdict
Three subagent audits reviewed the files that grew through the last rounds of patches. Their honest verdict: dense, not bloated; roughly 430 words of true redundancy came out with no rule lost, and every cut they flagged as removing compliance pressure was skipped. Highlights: approval recording now has one owner in visualize.md, the asset producer's crop ban went from three statements to the deliberate pair, its two transparency passages carried contradictory defaults (resolved toward true alpha first), and the 450-word medium-gate wall split into three paragraphs at zero cost. The producer also gained a mode-seam sentence so a sketch run cannot return an asset manifest. The reviewer's verdict is no longer soft: a derived disposition line (rebuild / fix / ship) opens every return, computed from the matrix rather than felt, recomputed after the verdict pass, and never softenable by the parent, who must report it verbatim. The second hamster-wheel run showed the parent inventing 'PASS WITH FIXES' over a matrix with MATERIAL contradicted on the focal element. Two additions from the same session's evidence: hard offset shadows outside a neobrutalist world join the craft floor's refusals (codex invents them without fail), and hookless harnesses must run detect.mjs once before the finish review, because codex has no hooks and the detector otherwise never sees the build at all. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
e54fd13a33 |
Close the line-art loophole and widen the rebuild directive
The second codex hamster-wheel run read the medium gate and still assigned a shaded, perspectived technical illustration to 'Authored SVG geometry': the world was an instruction booklet, so the affinity clause's 'diagrams' blessed the downgrade, and the page shipped as flat clipart against an illustration-grade comp. The gate now says style does not move the boundary: perspective, shading, figure drawing, or dense mechanical detail is illustration however line-drawn it looks, and authored SVG ends where drawing skill begins. The craft floor's sketchy-SVG rule carries the same sentence. The reviewer in that run built an honest matrix, MATERIAL contradicted on the focal element, and still emitted it as a fixable item the parent answered with CSS. The rebuild directive now fires when MATERIAL is contradicted on the focal element, not only when TYPE falls with it, and every asset-requiring fix must say 'produce: <region>' so it cannot be answered as a style tweak. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
f8a34335cb |
Let the surface own the sketch and comp aspect
A landscape frame was the silent default at every generation site, which is a composition error before the build starts for native apps and mobile-first surfaces. The sketch frame, the asset producer's single-sketch contract, and the comp instruction now state it: portrait at device viewport when the surface is a phone screen, landscape for desktop web. The decision page adapts in kind: portrait art overrides the 16/10 slot with its own exact ratio so nothing crops, and the deck narrows so portrait cards line up side by side. The --schema guidance tells the model the page handles either. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
4432b92bbb |
Harden the comp-to-build translation after the hamster-wheel failure
A codex greenfield build produced an excellent approved comp and then an abysmal page, and the reviewer approved it. The failure chain: the implementation inventory downgraded a photographic hero to 'silhouette in SVG' and sculpted panels to 'material finish: CSS'; the builder read 'no photography on hand' as a license to avoid photographic rendering; QA looked at one full-page thumbnail; the reviewer was spawned with the builder's forked history and then scored fix claims instead of pixels; and the output contract had no way to say 'rejected'. The fixes, stage by stage: - The inventory's medium column gets a gate: a human figure, product object, machinery, or lit material is raster whatever the stack, and such regions are regenerated cleanly at asset resolution with the comp and its embedded prompt as reference. Never cropped from the comp, whose effective resolution is reference grade; the asset producer's direct bucket closes the same hole. Dropping an image-native region is a user decision at the approval point. - Generated imagery is a material, not a claim: evidence rules bind assertions, never render fidelity. - The build thread's inspection becomes a region-by-region side-by-side against the comp at legible scale, never one full-page thumbnail. - The reviewer spawns fresh, never with forked history (fork_turns: 0 in codex), and gains a rejection lane: when TYPE, MATERIAL, and the focal element are all contradicted, the first material fix is a rebuild directive the parent surfaces to the user instead of patching. Verdict passes score recaptures only; the parent's fix narration is not evidence. - The verdict-loop ceiling softens: two rounds ends an unattended run, but an attended session puts the open-items table in front of the user and lets them fund another round; any round that resolves nothing stops the loop. - Comp approval joins the roll as skip-proof: question-tool errors fall back to the decision page, delegation is recorded in the brief and the sidecar and disclosed up front, and the reviewer treats comps with no recorded pick as a material finding. - Craft floor: system display faces (Impact, Arial Black) as an own-world display voice and unicode glyphs standing in for icon systems are named failures. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
47b875a7e3 |
One prompt carrier across every harness: embed-prompt.mjs
The prompt behind a generated image was recorded three different ways, a sidecar in the eval harness, nothing in the skill's API tool, nothing for native tools, so intent survived or vanished depending on where you ran. One dependency-free script now embeds the prompt inside the image itself, PNG tEXt or JPEG COM with a sidecar fallback for other formats, idempotent, and reads it back from any impeccable-generated file. The API tool embeds automatically; the prose directs every native-tool generation through it; copies between machines and harnesses keep their intent. Comps meanwhile are declared the build thread's own work, never delegated, and the comp-skeleton guidance now asks for the surface's actual regions instead of prescribing navs onto pages that have none. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
39532a65a2 |
Comps are pages not vignettes, and the prompt travels with the asset
Two findings from the first human-validated probe. The comps rendered as scene vignettes because the generation prompts led with the world's atmosphere; the model painted the fish market instead of the fish market's website. The comp guidance now demands the page's literal skeleton in the prompt, nav and its items, headline block, sections in order, footer, with a self-check: a render that could hang as a poster is not a comp. And generation context is part of the asset: the thread that wrote a prompt knows what the image contains and why, so build- critical imagery prefers the build thread, and subagent-produced assets must carry their prompts, via the tool's new sidecar or the manifest, read by the builder before composing. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
c4d22bb9dc |
TYPE and MATERIAL do not lapse when no comp exists
The failed gallery batch bound its seed, ran the reviewer, and still shipped CSS bevels imitating enamel: the matrix's material row was defined against the approved comp, and comp-less runs left it with no reference. The rows now fall back to the contract's OWN-WORLD and the world's real materials, with faked physicality contradicted on its face; imitation material is the single most reliable mark of machine-made design. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
8b6324d1b9 |
View every image by its workspace-relative path
A sandboxed harness rejected view_image on an absolute path to a mock the model had itself just produced under .impeccable/mocks/, killing the run. The relative-path rule existed only for downloaded quality-bar cards; it now covers every image the flow produces or references, in the comp round and in the asset producer's comparison step. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
6a7d75b6fe |
Bound the finish by verdict, not by count, and teach the matrix medium and type
The hard stop landed one step early: one review, one batched fix, one recapture, then done, with nobody ever judging whether the fixes reached the quality the findings named. A recapture measures positions; the model then presented mechanical confirmation as artistic success over a page whose display face, material, and hero legibility had all drifted from the approved comp. The finish now ends on a verdict: the recaptured screenshots go back to the same reviewer, which scores every material fix resolved, partial, or unresolved and names at most three regressions the batch introduced, no new hunt. Partial and unresolved fixes earn exactly one more round; two rounds is the ceiling, the second verdict ends the work whatever it says, and the final verdict table goes to the user as it stands, open items included. Three blindnesses from the same run close alongside. The matrix gains two mandatory rows: TYPE, where a display face of a different character is contradicted however the layout matches, and MATERIAL, where flat CSS standing in for painted, textured, or dimensional artwork is contradicted regardless of placement. And the Truth check now requires every produced asset visibly present in the screenshots, because a paper texture at 0.16 opacity is a compliance token, not a shipped material. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
09a33bc58b |
One sketch, one agent: retire the batch producer and its supervision
The batch producer was the clumsy piece: one subagent owning eight jobs needed heartbeat rules, reclaim windows, and a page full of fallbacks to survive its own opacity. The unit of work is now a single card. With parallel subagents, the set fans out one agent per card, up to four in flight, landing everything in roughly the time of one; a single-sketch agent has no planning phase and no batch to stall, so a failure costs one slot and its remedies fit one sentence: regenerate an empty slot when its agent returns, drop it when the user answers first. Without parallel subagents, the main thread generates in reading order after serving, and the harness's own generation display carries the progress. The page-side streaming is unchanged; it never cared who writes the files. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
0eb443d29b |
Bound the hand, greek the copy, and treat waiting as supervision
A codex field run dealt six challengers into an eight-sketch batch behind an opaque subagent, and the user stared at a page of shimmer asking whether anything was happening at all. Four fixes from that run. A hand now holds at most three challengers, the rest banked for re-rolls, so fairness within the hand stops multiplying into a queue. Sketches greek everything but the product's real name and one real headline, because an invented spec, price, or ship date in a sketch is a claim PRODUCT.md never made, and comps have solved this for a century. Sketch production follows the user's reading order with the first file doubling as the producer's heartbeat, and the parent's --wait loop checks the sketch directory each pass, reclaiming the batch inline when two minutes pass with nothing landed. And a failed --start now captures the daemon's stderr to a per-key log and names the sandbox as the usual suspect, instead of reporting only that failure occurred. The shimmer counts its elapsed seconds, and gives up at 150 instead of 300. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
d89ee5f87c |
Deal every card the same hand: anatomy, sketches, and the standing door
The decision page compared unlike things: the grounded direction was a wall of text beside curated catalog art, the catalog art read as a promise of the build, the weighing silently shrank the challenger set, and the standing exit hid in the footer under the cards it must not soften. Every card now shares one anatomy (thesis, palette chips, material tags, first viewport, case, risk), every dealt challenger is presented with the weighing written on it rather than applied to it, the catalog image rides picture-in-picture as labeled inspiration with the lightbox a click away, and canonCard renders the category standard as one honest, subordinate card. When image generation exists, each card declares a sketch slot the page polls: serve first, generate after, through one shared deliberately unfinished frame, so the comparison stays about direction instead of rendering luck. The asset producer takes the batch when subagents exist; the chosen sketch returns in ANSWER to seed at most one comp probe, and the comp round still renders its full set. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
f482d9405e |
Teach the reading-heavy subagents to write before the ceiling lands
Raising the reviewer's turn budget did not change its fate, only its reading: 43 tool uses instead of 22, still reaped mid-read with nothing written, because the SDK ends a run at max-turns without warning and the model never feels the deadline. The definitions now carry the deadline themselves: reading is an allowance, batch Reads per turn, take the decisive inputs first, sample instead of walking the tree, and write by mid-budget, naming what went unread. A review built from what you saw beats a perfect review that never arrives. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
c9213835e7 |
Review fidelity against the comp itself, not the builder's summary of it
A codex run turned an approved comp into a related second art direction and the finish reviewer passed it: the review anchored on the direction contract, a lossy abstraction the builder wrote, and every element that abstraction dropped passed silently. Four changes close that chain. The reviewer inventories the comp's salient elements before reading the contract and classifies each one (match, adaptation, missing, contradicted, added without approval), with adaptations citing the answer, brief, accessibility need, or product truth that forced them, and fidelity failures outranking craft in material_fixes. The visualize inventory gate records compositional commitments alongside asset media, since the 150-word contract cannot carry them. The north-star allowance now says what it permits: translation, never recomposition. And the finish sequence recaptures the same viewports once after the fix batch, so what the documenter records is what actually shipped. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
9e4990765f |
Give the reading-heavy subagents turn budgets that survive their inputs
A finish review reads the artifact, two full-page screenshots, the approved comp, the quality-bar cards, and the contract before it may write a word; at max-turns 12 the SDK reaps it mid-read and the parent receives the opening sentence as the whole review. Observed twice in a row (spawn and respawn) on the first real subagent run. The documenter reads at least as much, and the asset producer pays per asset. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
9c395bc484 |
Asset producer: codex notes as standalone blocks the compiler handles
compileProviderBlocks only processes standalone-line blocks, so the inline codex spans leaked literal tags into every provider's agent output, degraded fallbacks included. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
bb57be4243 |
Documenter subagent, reviewer handoff contract, asset gate
From the paired Opus and Codex manual-run analyses. DESIGN.md moves to the end of the flow and into a shipped documenter subagent that derives the system from the built artifact: a rulebook written before the build gets defended against reality, and a half-stable DESIGN.md hands the design-system detector an unstable target that buries the build in noise and invites laundering. The finish reviewer gains the handoff that failed three times live: the parent captures desktop and mobile screenshots and passes paths, the reviewer never attempts to render and names missing inputs in one line, the parent verifies the five-section return and respawns once on empty. Fidelity against the approved comp joins its checks; the card keeps commitment only. The comp ingredient inventory becomes a written gate with raster-by-default materials and no gradient-as-texture, comps persist under .impeccable/mocks, the degraded seed names the sandboxed-exec cause, and the finish line is explicit: a clean detector pass is not finished. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
253f8e510c |
Concept machinery: survive truncation, builds, and loud briefs
The release-gate audit traced four ways the roll's output was defeated downstream of a perfectly healthy seed. Gemini's harness keeps only the tail of tool output, so the header-only ASSIGNED INDEX never reached the model in 18 of 18 samples; the seed now restates the assignment and key at the end of its output. Astro strips frontmatter comments, so half the anthropic contracts vanished from built artifacts; the contract now must survive the production build as an HTML comment in emitted markup. A brief that paints its own picture (the album named Soft Cathedrals) converged every arm regardless of assigned index; its literal reading now joins the rut with at most one candidate. And Opus under 4.0.1 skipped the seed 42% of the time while hand-authoring plausible contracts; the finish reviewer now verifies FORM carries a corroborable seed key before any craft point. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
Paul BakausandClaude Fable 5
|
0bbb63b62a |
Ship the finish reviewer as a named subagent; ungate the asset producer
The eb686f36 session read the separate-reviewer rule and spawned nothing: an unnamed "separate agent" is an improvisation prompt, not an affordance. The skill now ships impeccable-finish-reviewer next to the asset producer: persistence first, ceiling against the card and comp second, contract promise by promise, truth; ordered material fixes back to the parent, no editing, no second detector. new-work names it so the finish step invokes a thing that exists. The asset producer was gated providers: codex, so Claude Code never shipped it; the gate is removed and its two codex-only workflow lines made provider-neutral with codex blocks. Dist rebuild still deferred for the running campaign. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
e8e3665142 |
Live mode: staged AI copy edits (#158)
* feat(live): manual text-edit panel + Astro inject + stale-lockfile reap Adds a manual text-edit popover under the live-mode bar so users can retype copy directly without going through generate. The footer's "Apply edits" button fires a manual_edits event; the server writes the changes back to source via the new live-edit.mjs deterministic file mutator. Mirrors the wrap+accept flow but skips variant generation. New scripts: - skill/scripts/live-edit.mjs: writes manual_edits back to source - skill/scripts/live-text-rows.js: browser walker that surfaces every pure-text descendant of the picked element as an editable row Touched scripts: - skill/scripts/live-browser.js: text panel UI, CONFIGURING state hook - skill/scripts/live-poll.mjs: manual_edits routing - skill/scripts/live-server.mjs: manual_edits endpoint + handler - skill/scripts/live-wrap.mjs: small adjustments to support the flow Docs + tests: - skill/reference/live.md: manual-edit section - tests/live-edit.test.mjs, tests/live-text-rows.test.mjs Also bundles two live-mode reliability fixes that surfaced during manual testing of the feature: 1. live-inject now emits is:inline when the inject target is a .astro file. Astro otherwise processes the <script> tag and rewrites src to its own bundled URL, so the literal live.js never loads. 2. readLiveServerInfo now probes the lockfile PID with kill(pid, 0) and unlinks the stale lock if dead. Previously a crashed helper left server.json with a dead PID and live-poll reported "Live server not running" forever. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(live): inline contenteditable text editing Replace the text-edit popover panel with inline contenteditable activation. When an element is picked in CONFIGURING, every pure-text descendant becomes contenteditable="true" directly on the page. Each blur-event fires a single-op manual_edits save to source. Esc restores original text and stays in CONFIGURING; successful save exits to PICKING. If Go is clicked while a save is in-flight, the save completes before generate fires. Deleted ~340 lines of panel UI (initTextPanel, openTextPanel, closeTextPanel, renderTextRow, buildTextFooter, etc.). Added enableInlineEdit, disableInlineEdit, onInlineBlur. Server contract unchanged; live-edit.mjs handles per-op saves as before. Tests: 186 pass. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(live): hide annotation overlay during inline edit Annotation overlay's click handler was intercepting clicks on contenteditable text elements. Hide the overlay when inline-edit is enabled to allow text selection and editing. Restore it when exiting inline-edit (if still in CONFIGURING). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(live): edit content badge mode with batched saves Replace automatic inline contenteditable on element pick with an explicit "Edit content" badge. The badge appears at the element's top-right corner when an element is picked. Clicking the badge enters a new EDITING state where: - The contextual bar hides - The annotation overlay hides - The badge morphs to show Cancel + Apply buttons - Text descendants become contenteditable inline Edits are held in memory (input event tracking) until Apply is clicked, which fires a single batched manual_edits event with all ops. Cancel discards drafts without saving. This eliminates the annotation overlay interference that prevented clicking on text elements. The EDITING state integrates with the main state machine and handles all exits (Esc, click-outside, teardown) cleanly. All 186 tests pass. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(live): use row.el.tagName for tag in applyEditing op The applyEditing function was trying to use row.tag which doesn't exist on the row object. The tag should be the tagName of the text element itself (row.el.tagName.toLowerCase()). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(live): Edit content badge styling + auto-focus + separate buttons - Edit content button now matches Go button styling (BP.accent background, BP.mark text, FONT, transitions, hover effects) - Auto-focus first editable element when entering editing mode (50ms timeout) - Separate Cancel and Apply buttons with 8px gap (no divider) - Cancel uses muted styling (BP.hairline background, BP.textDim text) - Apply keeps brand accent styling - Remove all focus rings and outlines on edit badge buttons (no blue ring/outline in EDITING mode) Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com> * feat(live): Subtle button UI + cursor positioning + better copy - Change badge buttons to use impeccable-button aesthetic (ink background, surface text, hover to accent) - Removes aggressive styling conflict with Go button - No animations; simple 150ms background transition - Matches site design language (padding 0.625rem 1.5rem, 0.8125rem font, letter-spacing 0.03em) - Shorter, clearer button copy: "Edit" instead of "Edit content", "Save" instead of "Apply" - Fix cursor positioning: cursor now appears at END of text, not beginning - Use Selection API to collapse cursor to end of contenteditable element - Improves UX for immediate continuation of text - Update live.md documentation to reflect new button labels Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com> * fix(live): Use site design system colors for edit badge buttons - Edit/Save buttons: oklch(10% 0 0) background → oklch(60% 0.25 350) on hover - Cancel button: oklch(55% 0 0) background → oklch(65% 0 0) on hover - All buttons: 6px border-radius (matches Go button), oklch(98% 0 0) text - Smooth transition: 0.3s cubic-bezier(0.16, 1, 0.3, 1) (--ease-out) - Uses site color palette instead of live-overlay constants Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com> * fix(live): Match slop-callout style for edit badge buttons - Use exact .slop-callout aesthetic: paper background, accent border + text, uppercase 10px (0.625rem) - 600 weight, 0.06em letter-spacing, 4px 8px padding, 6px border-radius - Box-shadow: 0 2px 8px rgba(0,0,0,0.1) matches site callouts - Hover: inverts to filled background (accent fill, paper text) - Cancel uses ash color variant for muted state, Save uses accent - Smooth 0.3s cubic-bezier(0.16, 1, 0.3, 1) transition on background and color Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(live): Pill-shaped edit badge buttons, 2px padding, no uppercase - Border-radius: 999px (pill shape) - Padding: 2px 8px (more compact) - Removed text-transform: uppercase Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(live): Cancel button uses mist border + ash text - Border: 1px solid oklch(92% 0 0) (--color-mist) - Color: oklch(55% 0 0) (--color-ash) - Hover: inverts to ash background with paper text Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(live): Remove blue focus outline from contenteditable elements in EDITING mode - Add inline outline: none on each row's element when contenteditable activates - Inject [data-impeccable-editable] CSS rule to override browser default focus ring - Use !important to win against site styles that re-apply focus outlines - Cleanup restores outline/data-attribute on disable Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(live): Decouple manual edits from agent/poll pipeline Manual text edits now POST directly to a new /manual-edit endpoint that runs live-edit.mjs synchronously and returns the result. The event is never enqueued, never reaches the poll loop, never reaches the agent. Why: every Save was costing an LLM turn. The poll script would dequeue the manual_edits event, run live-edit.mjs deterministically, post a completion ack, then print the event JSON to stdout. The Claude agent would read that output and decide "loop and re-poll". Zero real work for the agent but every Save burned context. Changes: - live-server.mjs: new POST /manual-edit handler that runs live-edit.mjs synchronously and returns the result. Does not enqueue, does not log to session store. Defense-in-depth: /events rejects manual_edits. - live-browser.js: applyEditing() POSTs to /manual-edit instead of sendEvent({type: 'manual_edits'}). - live-poll.mjs: removed manual_edits handler branch (dead code now). - reference/live.md: removed "Handle manual_edits" section; replaced with a one-line note that manual edits are server-direct. The HMR-triggered page reload remains (dev server detects source file change) but that is a separate dev-server behavior, not our pipeline. resumeSession() already restores variants and selection after reload. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(live): Stash manual edits server-side; commit via AI on request Decouples manual-edit Save from source file writes. Save now stashes to .impeccable/live/pending-manual-edits.json with no HMR refresh. The user explicitly asks the AI to commit when ready. Why: even with the prior /manual-edit fix, every Save still wrote to source and triggered the dev server's HMR/full reload. The page flash was the actual user pain. Now there's zero source touch on Save, and the user controls when the dev server reloads. Server (live-server.mjs): - /manual-edit-stash POST: append to buffer file. Returns {ok, pendingCount, totalCount, perPage}. - /manual-edit-stash GET: query counts by page for counter UI. - /manual-edit-discard POST: drop entries (all if no pageUrl). - Old /manual-edit returns 410 Gone (defense in depth). - Buffer ops merge by (pageUrl, ref): keep first originalText, update newText. CLIs: - live-commit-manual-edits.mjs: read buffer, shell out to live-edit.mjs per entry, truncate succeeded entries, surface failures. - live-discard-manual-edits.mjs: truncate buffer (optionally scoped by page). - Both take optional --page-url=<url>. Browser (live-browser.js): - applyEditing() POSTs to /manual-edit-stash, no source write. - Pending pill (• N staged) + trash icon next to Exit in global bar. - One-time onboarding toast on first Save: "Saved. Tell the AI to commit when ready." - Counter persists across reloads via GET /manual-edit-stash on init. - Trash icon: confirm dialog scoped to current page, then POST /manual-edit-discard. Variant pipeline interaction: - live-wrap.mjs: when wrapping an element, apply pending manual edits to the source range so the wrap block's "original" variant reflects the user's edited DOM (their pre-Go view), not the raw source. - live-accept.mjs: after accept writes the variant to source, scrub buffer ops whose originalText no longer appears in that file. The accept embodies the manual edit; the pending op is consumed. - Variant discard does NOT touch the buffer. Reference docs: - reference/live.md: full commit/discard contract, trigger guidance (narrow action-verb intent), do-not-auto-commit rule. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(live): Staged-edits pill becomes an "Apply" button Click the "• N staged" pill → confirm dialog "Apply N staged edits to source? The page will reload." → POST /manual-edit-commit on the server, which shells out to live-commit-manual-edits.mjs. Same path the AI uses, just triggered from the overlay. Trash icon stays for discard. The AI-driven commit path also stays (useful for inspecting failures or scripting). The pill is now the primary apply affordance because it removes the chat-context-switch for the common case. Pill styling: pointer cursor, accent border + text at rest, fills on hover (accent bg, paper text). Tooltip: "Click to apply staged edits to source". First-save toast updated: "Saved. Click the 'staged' badge to apply, or ask the AI." Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * chore(live): gitignore pending-manual-edits.json runtime buffer Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: drop stray site/ test edits from PR Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(live): Pill label reads "Apply N staged" Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(live): Manual edit ops use the leaf element's locator, not parent's Multi-row inline editing captures each contenteditable leaf (row.el) but the op was being built with selectedElement.id / classList — i.e. the parent card, not the editable text node. live-edit.mjs then searched source for the parent's class on the leaf's tag (e.g. <span class= "foundation-card">), found nothing, and silently failed. Use row.el's own id / classList instead. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(live): Climb to nearest classed ancestor when leaf has no locator A bare <em>/<strong>/etc. with no id or class produced ops the CLI rejected with insufficient_locator. Prefer the leaf's own id/class; if neither exists, walk up to the nearest ancestor with one and adopt its tag + locator. Text-replace still works because the CLI narrows by originalText inside the matched element's source range. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(live): Make mixed-content paragraphs editable The text-rows walker skips elements with mixed children (text + element + text), so paragraphs like "Some text <code>x</code> more text" or "Body text · <a>link</a>" exposed zero rows for the surrounding copy. At edit time, wrap each non-whitespace direct text-node child in a marker span so the walker emits a row for it. Unwrap on save/cancel. The locator climbs to the parent's class as before, and live-edit narrows by originalText inside that parent's source range. hasTextRows now uses a lightweight subtree check that matches the new wrap+walk path so the edit affordance shows up on mixed-content elements. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(live): Address Cursor Bugbot findings (CB-2 through CB-6) CB-2 - Escape reverted DOM text but inlineEditDrafts retained the pre-revert value; clicking Apply afterwards committed the undone edit. Clear the draft entry when restoring innerText. CB-3 - The scrub gate !result.handled || result.handled !== false was a tautology that ran the scrub regardless of accept outcome. Use the intended result.handled !== false. CB-4 - The buffer-aware "original" content step in live-wrap iterated every entry in the buffer with no pageUrl filter, so an edit on /a could leak into a wrap call on /b. Add --page-url to the CLI; filter by it; skip the buffer-aware step entirely when omitted. live.md updated. CB-5 - removeEntries returned entry count while truncateBuffer returned op count, causing the discard CLI and HTTP endpoint to report mixed units. Make removeEntries return ops removed. CB-6 - applyTextReplace used string truthiness to gate prepending content above the edit, which silently dropped a leading empty line when the file started with '\n'. Gate on the line index instead, and mirror the fix on the trailing-empty-line side. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(live): A3+A4 data-integrity guards, A6 test coverage A3 — applyTextReplace refuses with text_ambiguous_in_block when originalText appears more than once in the matched element block. Refusing is safer than picking the first indexOf hit when we can't tell which leaf the user edited; user can rephrase one occurrence. A4 — newText is rejected if it contains <, >, {, }, or a backtick. Two layers: server-side validator in /manual-edit-stash returns 400, CLI-side guard in applyTextReplace returns invalid_chars_in_newText. Browser surfaces the specific reason via toast. The shared char list lives in live-edit.mjs (validateNewTextChars). reference/live.md documents the rule. A6 — New test files cover the orchestration gap: - live-manual-edits-buffer.test.mjs (17 tests across read/stage/ remove/find/count/truncate; pins removeEntries returns OPS count) - live-wrap-buffer-aware.test.mjs (3 tests; CB-4 regression test) - live-commit-manual-edits.test.mjs (4 tests; partial-failure, --page-url scope, no_pending_edits) - live-discard-manual-edits.test.mjs (3 tests; CB-5 unit consistency) - live-accept-scrub.test.mjs (4 tests; keep/drop/prune) Plus 2 new cases in live-edit.test.mjs for A3 and A4. Side-effect refactors: - scrubManualEditsAgainstFile accepts cwd for unit-testing and is exported. - Failed-op entries in live-edit.mjs now propagate forbidden and occurrences fields so callers can surface specifics. 41 tests across the 6 affected files pass; full suite green at 186/186. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: drop .claude/pr-review.md from PR Local review notes belong in the working tree, not the PR diff. Kept in the file system; just untracked. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: drop stray site/ test edits from PR (round 2) Live-inject script tag and the "Impeccable Works!" / "WHAT'S INCLUDED IN THE BOX" / "Wow Impeccable. ---- " strings were test edits that slipped back into the branch. Restore both files to match main. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(live): Disable Edit badge while variants are generating Clicking Edit during GENERATING would open inline text editing on the same DOM region the variant wrapper is about to land in, racing the HMR and the mutation observer. The badge now switches to an 'idle-disabled' rendering (ash + mist, not-allowed cursor, disabled attribute, tooltip) the moment state transitions into GENERATING. Returns to 'idle' on the normal CONFIGURING re-entry paths. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(live): live-wrap refuses without --page-url when buffer has pending edits When a manual edit is staged ("Impeccable Works!") but not yet committed, the buffer holds the user's edited DOM while source still has the un- edited text ("Impeccable"). live-wrap's buffer-aware step exists to rewrite the wrap block's <div data-impeccable-variant="original"> to match the staged DOM, but per CB-4 it is gated by --page-url. When the agent invoking live-wrap omits --page-url, the buffer-aware step silently no-op'd and the variant authoring saw stale source — the user's manual edit appeared lost. Make the silent no-op a loud error: when buffer.entries.length > 0 and --page-url is missing, exit 1 with { error: 'missing_page_url_with_pending_edits', pendingEntries, hint }. Empty buffer = no risk = no requirement, so existing flows without pending edits keep working. Updated reference/live.md to flag --page-url as required when the buffer has entries. Added regression test in live-wrap-buffer-aware.test.mjs. live-wrap.test.mjs gained a buffer- clear hook so any leftover .impeccable/live/pending-manual-edits.json from local dev doesn't trip the new check. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * change back * chore: drop stray site/ test edits from PR (round 3) Live-inject script tag in Base.astro slipped back in via git add -A while a local live server was running. Restore both site/ files to main. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * Fix live manual edit staging * Rename live edit copy badge * Use sentence case for live edit copy badge * Move copy edit apply control outside live bar * Improve live copy edit apply flow * Clean up live copy edit AI apply flow * Polish live copy edit docs and toast * Fix staged copy edit review issues * Fix CI jsdom dependency * Fix Cursor Bot live edit findings * Fix remaining live edit review issues * Fix Bugbot staged edit edge cases * Fix latest Bugbot live edit edges * Fix remaining Bugbot wrap and discard issues * Fix live copy edit safety contracts * Fix copy edit rollback coverage * Fix live manual copy edit apply flow * Adjust live pending dock offset * feat(live): route manual-edit Apply through the chat agent Make the staged copy-edit Apply work when no CLI AI runner is authenticated by routing the batch through the active chat session, and surface runner failures clearly instead of opaque exit codes. - live-poll: add --reply --data '<json>' so the chat agent can return a structured manual_edit_apply result (the documented flag was missing, so the server resolved with an empty object) - live-server: manual_edit_apply event + deferred map, chat-vs-subprocess dispatch in /manual-edit-commit, resolve the deferred from the ack - live-copy-edit-agent: chat provider, extractRunnerErrorMessage and commandAuthed pre-flight, diagnostic describeNoProviderError; drop the stale CLAUDE_CODE_SIMPLE and --no-session-persistence flags so headless CLAUDE_CODE_OAUTH_TOKEN auth works - live-browser: clear pendingApplyInFlight on commit_done and add a watchdog so a missed signal can no longer freeze element picking - reference/live.md: tight Handle manual_edit_apply handler plus a separate diagnostics reference section; advertise the event in the opening contract and dispatch table Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * Add live manual edit apply coverage * Fix manual edit apply review issues * Fix manual edit review follow-ups * Fix manual apply poll acknowledgements * Fix manual apply failed-entry rollback * Clarify manual apply LLM prompt * Fix stale manual apply discard events * Fix manual apply dynamic source edits * Fix large manual apply chunks * Clarify manual edit apply is first-class work * Clarify manual apply resume flow * Compact live manual apply evidence * Reject malformed manual apply replies * Recover legacy manual apply summaries * Fix Astro live script injection * Add live manual edit apply coverage * Slim live manual apply flow * Slim manual edit test dependencies * Stabilize real browser LLM smoke * Generalize manual edit LLM prompt examples * Remove retired live edit wrapper * Inline live text row walker * Slim manual edit prompts * Drop AGENTS doc churn * Stabilize live manual apply prompts * Stabilize manual apply visible Haiku flow * Add hard framework manual edit coverage * Stabilize manual edit LLM retries * Fix manual apply transaction rollback * Fix live shader text capture * Clean up manual apply runtime artifacts * Fix live manual edit apply reliability * Clean up manual apply coverage * Slim manual apply test cleanup * Fix manual edit prompt contract test * Align manual edit cancel hover * Fix live loading shader capture * Fix manual apply review findings * Restore live e2e tests for CI * Fix live loading shader halftone * Tune live loading shader dots * Restore main live shader behavior * Fix manual apply review findings * Fix manual apply bot follow-ups * Clarify manual apply rollback changes * Fix manual apply state naming * Address PR review cleanup * Fix manual apply review follow-ups * Fix multiline manual apply verification * Restore inline drafts when hiding live bar --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
e7e923c4ef |
Skill + craft cleanup, detector hardening, native subagent pipeline (#152)
* skill: drop quality tiers, keep the real brand-craft guardrails Codex's craft/brand pass introduced fast/ship/showpiece "quality bars" plus brand-specific build gates, asset ledgers, sub-agent review, and self-graded fallback labels. In practice those tiers became escape hatches rather than craft pressure: the final output should always be 10/10, and the real decision points are splashiness and maximalism, not quality. Removed: - All quality-bar / showpiece / fast / ship framing in shape.md and craft.md - Standalone Brand Direction (#4) and Asset Requirements (#10) sections in shape's brief; renumbered back to 1-10 - The Brand hard rules section in brand.md (folded its real prohibitions into the existing Imagery and Brand bans sections) - Brand-specific build-gate item, mock-fidelity bullet, production-bar bullet, present-step bullet in craft.md - Asset ledger ceremony in craft Step 4 - Review-only sub-agents and "self-reviewed fallback, not independently validated" machinery in craft.md and polish.md - The For brand surfaces, assess hard failures subsection in polish.md and the brand checklist row - tests/brand-showpiece-reference.test.mjs (and its package.json wiring) Kept (the real nuggets): - Asset-substitution prohibition: image-led briefs ship real/generated assets or canvas/SVG/WebGL, not generic CSS panels, cards, bullets, or copy - Repeated tiny uppercase tracked kicker labels as a brand ban - Detector/QA output is defect evidence only, never proof of quality - "What visual assets are real content here?" discovery question - Inspect each major section individually for brand and long-form work - repeated-section-kickers detection rule + fixture - CLI improvements (JSON to stdout, -json/-fast aliases, severity field) - critique.md: npx impeccable detect --json fix Harness output dirs refreshed via bun run build. Full test suite (186) passes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * skill: strip gate ceremony; require shape pause; allow compact briefs The setup gate table and IMPECCABLE_PREFLIGHT banner pushed every craft run through ritual restatement (PRODUCT.md → original prompt → round 1 → round 2 → 70-line "confirmed brief" → critique → summary, all saying the same thing). Replaced with imperative prose that still demands the same work but skips the user-facing telemetry. Specifically: SKILL.md - Drop the Setup gate table and IMPECCABLE_PREFLIGHT banner. - Keep the imperative steps explicitly: load context, identify register and load brand.md or product.md, AND load the matching command reference (craft.md / shape.md / etc.) when a sub-command is invoked. The command-reference step is non-negotiable; without craft.md loaded the agent skips the shape-and-confirm pause. craft.md - Drop the Build Gate / Craft Contract formal sections; replace with one paragraph stating prerequisites. - Step 1 explicitly requires ending the response after presenting the shape output; the user must confirm before any code lands. Allows a compact 3-5 bullet brief when the prompt + PRODUCT.md already pin direction (full 10-section structure reserved for genuinely ambiguous tasks). - Step 3 image gate skips silently when image generation isn't natively available; no user-facing announcement. - Step 6 explicitly legitimizes "first pass clean, shipping" as a valid endpoint and bans inventing fake defects to demonstrate iteration. shape.md - Cap discovery at 1 round by default; second round only when first leaves material gaps. - Adds an "assert-then-confirm, not menu-with-escape" rule: when PRODUCT.md and the prompt make one option obvious, name it and ask for confirm or override instead of enumerating "Restrained / Committed / Or something else?" as a real choice. - Phase 2 brief has two forms now: compact (default for clear briefs) and full structured (genuinely ambiguous). Open Questions can't double as leading-with-Recommend; if you'd write "Recommend: X", decide X. - Image gate same as craft.md. Validated end-to-end with a Haiku skill-on observability run: agent loads craft.md plus the brief's recommended implementation refs, pauses for one productive question (accent color, trace fidelity, CTA), and ships an artifact with zero side-tab violations vs. the original v1 baseline. Cost trades up modestly for that quality. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * craft.md Step 6: Reading the screenshot is the inspection, not taking it A v4 eval run took 4 targeted screenshots (hero, mobile, tablet, query-section) and then never Read any of them back. The agent treated browser_screenshot itself as "I inspected" and shipped without the multimodal feedback loop ever closing. Detector caught the resulting slop (5+ side-tab violations) on adjacent runs that did the same thing. Step 6 now spells out the pattern explicitly: take the screenshot, then Read the resulting PNG so its image content enters the conversation as multimodal input, then critique what you actually see in the image. With a check: "if your critique could have been written without looking at the image, you didn't look at the image." Validated with v5b: agent took 6 screenshots, Read all 6 back, and shipped with zero detector findings (vs the previous greenfield runs that hit 1-12 findings each). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * craft + brand: framework foundation, build-pipeline respect, image verification Three closely-linked additions surfaced by an eval-harness session investigating why the agent always shipped flat single-file HTML and zero imagery on greenfield brand briefs. 1. craft.md gains a new Step 0 "Project Foundation" before Shape. Detects existing framework / component library / icon set and uses what's there. Greenfield: ask the user via AskUserQuestion with sensible defaults framed by the brief (Astro for content/ brand sites, SvelteKit/Next/Nuxt for app surfaces, single index.html only for one-shot demos). Skipping the framework decision and writing flat HTML "to satisfy the spec" produces work that reads as a 2018 prototype regardless of visual quality. 2. craft.md Step 5 production bar gains two bullets: - Respect the build pipeline. Edit source files and run the project's `npm run build`; do not write to build/ / dist/ / .next/ directly with cat/heredoc/Bash redirects. Bypassing the pipeline skips asset hashing, image optimization, code splitting, and CSS extraction. - Verify external image URLs before referencing them. Use an image-search MCP, web-fetch tool, or browser if available; guessed photo IDs ship as broken-image placeholders. 3. brand.md "Imagery" section: - Generalizes the Unsplash URL guidance to "verify URLs before referencing them" with a hierarchy: image-search MCP > web-fetch > confidence-restricted manual selection > fewer photos. - Tightens the tech/dev-tool exception. Old line "zero imagery can be correct" gave models a permission slip. New framing keeps the underlying truth (typography + code + diagrams primarily carry voice) but raises the floor: imagery still earns its place when it serves the brief, and skipping it requires naming the typographic/diagrammatic move that's carrying the visual weight instead. "Zero imagery is the failure mode of laziness, not restraint." Eval-harness corpus that prompted this: 19/19 brand landing tasks shipped 0 images each, including ones where Opus had taste enough to break the dev-tool color default lane. The skill needs to teach both halves of the decision; the harness shouldn't have to nudge. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * detector: body-text-viewport-edge rule + OKLCH/var-resolution + anchor-inherit FP fixes New rule: body-text-viewport-edge flags body paragraphs that render flush against the left/right viewport edges (no container padding). Tested via the new tests/fixtures/antipatterns/body-text-viewport-edge.html fixture (3 flag cases, 5 pass cases) and the test in detect-antipatterns-browser. False-positive class fixes — all jsdom-mode only (real browsers resolve the cascade correctly so these gates stay inert there). Five related gaps that compounded into ~14× spurious contrast findings on Tailwind v4 pages with OKLCH color tokens: • OKLCH parser. jsdom returns the literal "oklch(...)" string from getComputedStyle; the detector now converts to sRGB via Björn Ottosson's matrices. Handles Tailwind v4's compact minified form "oklch(21.5%.02 50)" (no space after %). • var() resolution. resolveBackground + checkElementColors now accept the existing customPropMap and parse `var(--color-paper)` etc. as proper RGB via the new parseColorResolved helper. • bg-color before bg-image. The old order bailed on any gradient ancestor before checking for a solid background-color underneath, causing the body's decorative paper-grain gradient to be measured against instead of the page's actual `bg-paper` cream. • body/html-level gradient → white fallback. When the only opaque ancestor we can read is body/html with a gradient overlay (and jsdom can't decompose `background: var(--paper) gradient` to extract the solid color), return white instead of falling through to resolveGradientStops — which was picking up paper-grain noise colors and using them as the bg. • Anchor-inherit workaround for jsdom :link UA specificity. Tailwind v4's preflight declares `a { color: inherit }` (0,0,1). jsdom's UA stylesheet has `:link { color: blue }` at (0,1,1) and wins the cascade. Real Chrome wraps :link in :where() (0,0,0) so the page rule wins. When the page declares the inherit rule AND we see jsdom's default `rgb(0,0,238)` on an anchor, walk to the nearest non-anchor ancestor and use its color. • Alpha-fallback safety gate. When text has alpha<1 AND we couldn't find an opaque ancestor (effectiveBg null), skip the contrast finding. Covers any remaining FP class the deeper fixes miss. Verified end-to-end against an Opus iter-1 artifact on Tailwind v4 with 14 cream/cream FPs + 2 blue-link UA FPs before; 0 findings after, while the color.html fixture's 12 real low-contrast cases continue to flag (verified via direct detectHtml calls). cli/engine/detect-antipatterns-browser.js is the generated browser distribution — regenerated from .mjs via scripts/build-browser-detector.js (no manual edits to the generated file). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * craft.md: tighten verbose passages, de-codex Step 6, cut redundancies Cumulative reduction: 218 → 155 lines (-29%). Step 0: drop the "Why this matters" paragraph at the end. The body of Step 0 already makes the framework-pick point; the paragraph just re-explains it with extra rhetoric. Step 1: replace the 4-sentence "you must end your response" block with a single line. The original said the same thing three different ways. Step 3: trim the conditional / defensive scaffolding (Purpose subsection, "do not skip because the eventual UI is semantic..." paragraph, duplicated approval-loop guidance). Mock fidelity inventory preserved. Step 4: drop the "keep UI text semantic" sentence; it duplicates Step 5's "Semantic first" rule. The rasterized-vs-semantic decision rule stays. Step 5: tighten each production-bar bullet to bold-lead + specifics format. All 15 rules preserved (real content, mock ingredients, semantic first, spacing/alignment, typography, state coverage, interaction quality, icon set, build pipeline, image URL verification, optimized imagery, premium motion, maintainability, technical cleanliness, ask-when-uncertain). Step 6: rewrite around "look at what you built like a designer would — your eyes are whatever the harness gives you." Drops Codex-specific "In Codex, use browser-use" bias. Drops the verbose 3-step Read pattern (condensed to one sentence). Drops the 1-8 numbered checklist (replaced by a tight paragraph). Keeps the load-bearing rules: read the PNG, don't fabricate iteration, mock fidelity reference, exit bar = studio defensibility. Step 7: drop the closing "Iterate based on feedback. Good design is rarely right on the first pass" preachy filler. All em-dashes converted to semicolons / colons / periods to satisfy the skill prose validator. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * build: native subagent pipeline + Codex-only asset producer Adds an agent cross-compile pipeline alongside the existing skill pipeline. Sources live at skill/agents/*.md; providers that declare agentFormat (codex-toml, claude-md) emit native subagent files. An optional providers: <list> field on an agent gates which harnesses get a copy; default (no field) ships everywhere. The impeccable-asset-producer agent is opt-in to Codex only. It's useful for Codex's native image generation path and is untested elsewhere; Claude has no native image gen anyway. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * brand: inverse-test + cultural-symbol palette guardrail Two additions to the brand register reference: - Inverse slop test: describe the page the way a competitor would describe theirs. If that sentence fits the modal landing page in the category, restart. - Palette guardrail: when a cultural-symbol palette is the obvious pull, reach past it. Let cultural reading come from typography, imagery, and copy. Harness mirrors regenerated; some also catch up to the image- verification paragraph from |