probe-compking-sol-4 (evals) executed its staged inventory faithfully and
still lost the comp: the brief had already recorded the comp's materials
down (low-contrast textures, a 70-path lake against the comp's hundreds,
a sculpted plate as flat CSS), and the build thread never loads
visualize.md, so nothing told it the comp wins that disagreement. The
comp-is-king block now says the record gets corrected upward, that the
comparison runs against the freshly reopened comp rather than memory,
and that a texture under a near-opaque wash is not shipped material.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Phase one is near-pixel-perfect reproduction at the comp's breakpoint,
with exactly three concessions (closest font, icons unless a library
was chosen, genuine comp defects); the overlap comparison is the
authority because models systematically believe their code recreation
succeeded when it did not, and regions that keep losing the comparison
ship as composited rendered assets instead. Phase two brings the
reproduction to life (interaction, motion, responsiveness), and
anything beyond the comp inherits the recorded system, never invented
container chrome.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
context.mjs now probes cwebp/sips/magick/ffmpeg once (which/where per
OS) and prints IMAGE_TOOLS, replacing macOS-specific prose; the
IMAGE_GEN_AVAILABLE directive leads with the harness-native tool so a
present OpenAI key stops reading as an instruction to bill it; and the
sandboxed board-start guidance sheds codex vocabulary.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Sandboxed shells cannot bind the board's port; every codex session paid
one failed start before retrying escalated.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three lessons from the Tortuga containment-map run. The first rebuild
directive now executes immediately, informing the user instead of
asking permission to fix a failure; consultation waits for a second
rebuild verdict or user-approved content at risk. The asset producer
spawns on every subagent-capable run even when produce looks empty,
because its manifest is the independent check on the inventory's media
and the skipped spawn marks every all-CSS failure to date. And the
inventory now opens by reading the comp as a design system (component
grammar, corners, line weights, elevation, type ramp), because the
sections the comp does not show get built from that record, and without
it the fallback is the stock kit: square boxes, 1px grids, bentos, hard
shadows.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Tortuga run showed the finish machinery working end to end, roll,
challengers, approved comp, honest labels, independent reviewer, an
earned rebuild verdict at the user checkpoint, and still cost the user
a full build to learn the hero undersold the comp: a glyph storm at a
tenth of the approved density under type half as compressed. Both
misses were visible the moment the first viewport rendered.
Two cheap gates front-load that discovery. The build section gains a
hero checkpoint: capture the first viewport and set it beside the
comp's before any later section, judging scale and density as
quantities. The inventory gains the same quantitative discipline:
field and texture regions record density and coverage, and TYPE rows
name the compression class and render one headline word against the
comp before anything is built on the face.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The composition pool (stagings) is not ready: too thin to help, and its
draws crowd the decision it rides along with. concept-seed.mjs stops
rendering the staging block by default; IMPECCABLE_COMPOSITIONS=1
re-enables it for catalog development, and the draw machinery,
rating-weighted selection, and mode scoping stay intact and tested for
its return. new-work.md drops the dress-the-staging-challengers
instruction and the FORM contract's staging clauses; the surface-scope
roll still assigns which of the model's own structures gets built.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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>
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>
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>
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>
The github and cursor providers previously received only the generated
degraded/ inline fallbacks. Both harnesses support real custom subagents,
so the build now emits them from the same skill/agents/ source:
- GitHub Copilot: .github/agents/impeccable-<role>.agent.md with portable
frontmatter only (name + description; omitting tools grants all tools,
and Copilot has no documented model/effort/max-turns equivalents).
- Cursor: .cursor/agents/impeccable-<role>.md with name, description,
model: inherit, is_background: false, and readonly derived from the
agent's tool list (true only for the finish reviewer, which declares
neither Write nor Edit). effort/max-turns are skipped because Cursor's
effort option requires an explicit model id.
Agent bodies now also resolve {{scripts_path}} and strip rule markers in
the shared agentFormat pipeline, which fixes the previously unresolved
placeholder in the emitted Claude asset-producer agent.
The CLI installer places agents per scope: project installs write
<repo>/.github/agents/ and <repo>/.cursor/agents/; user-level installs
write ~/.copilot/agents/ (Copilot's user dir, not ~/.github/) and
~/.cursor/agents/, overwriting stale impeccable-* copies. Because
Copilot lets user-level agents shadow same-named project ones, a project
install warns when shadowing copies exist; Cursor gives project agents
precedence, so no warning there.
new-work.md and visualize.md extend their harness-naming clauses with
the Cursor and Copilot invocations. The degraded/ fallbacks keep
shipping for surfaces where the model still fails to delegate.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Transcript archaeology on the failed gallery batch split the binding
break three ways. One model authored a complete, correct contract that
Astro then erased: the compiler strips a slot's leading comment while
keeping deeper ones, so the contract now belongs to the root layout's
body as its first child, and the first production build gets grepped
for the seed key, because a contract the build erased is a contract
nobody can audit. Another model simply skipped the roll and built the
exact category default the seed exists to refuse; the roll step now
states outright that it has no substitute and no skip condition. The
third failure was the worker watchdog, fixed separately.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two probe runs on two different harnesses built complete pages and
declared done without ever entering the finish sequence: the reference
was read once near turn four and the finish choreography had fallen out
of attention thirty turns later. The one text a model rereads on every
edit is its own artifact, so the direction contract now closes with a
FINISH line naming the exit condition verbatim: unreviewed and
undocumented is unfinished; this build ends with the finish review, the
verdict, and DESIGN.md. A page that looks complete with that line
undischarged is not done, it is abandoned at the finish line.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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>
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>
Field data: the first image of a real batch took ninety seconds and the
page's 150-second fallback then silently promoted catalog art to full
bleed, unlabeled, which is exactly the this-is-your-design misread the
picture-in-picture treatment exists to prevent. The policy is now
patience while there is progress: a slot shows its inspiration only
after waiting four minutes with nothing landing anywhere on the page
for four minutes, the stand-in is dimmed and labeled 'inspiration ·
sketch pending', and polling continues so the real sketch still swaps
in whenever it arrives. Slots with no inspiration keep the honest
elapsed shimmer instead of folding. The parent's reclaim rule matches:
files landing steadily is health at any pace, and only total silence,
no first file in three minutes, takes the batch back inline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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>
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>
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>
Harnesses with no subagent capability now run each role inline from the
same single source. The build emits reference/degraded/<role>.md for every
agent in skill/agents/ (role name is the agent name minus the impeccable-
prefix), stripping frontmatter and prepending the inline-substitution
preamble. These pass through the same provider-block compilation and
placeholder replacement as ordinary reference files, so <codex> blocks and
{{placeholders}} resolve per target, and they land in the committed harness
dirs on build:release like every reference file.
Repoint the three capability-first fallback sites in the prose at the
generated files: new-work.md reviewer and documenter fallbacks, and
visualize.md asset-producer fallback.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Probe attribution on Opus 5 showed the screenshot bound working (42
to 16) while the real burner ran free: five rounds of node -e
micro-edits, eight rebuilds, and inline defect hunts absorbed the
reviewer's and documenter's jobs until the turn cap killed the run
mid-hunt. The two-round ceiling now names scans, micro-edits, and
rebuilds; after the second round the build thread stops polishing and
ships the rest through the reviewer (one batched fix pass, one
rebuild, stop) and the documenter.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Opus 5 turned the iterate-with-screenshots-until-it-meets-the-bar
instruction into 42 screenshot trips and 150 tool calls per build,
about forty dollars of cache churn a page, before ever reaching the
reviewer. Verification now batches: one desktop-and-mobile round after
the full build, fixes applied together, one confirming round, ceiling
two. Craft-floor's checks share those renders instead of earning
separate trips; per-tweak iteration is live mode's channel.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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>
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's call: the build-exhaustion failure only exists inside eval
workers with hard turn budgets no real harness exposes, and the clause
doubled as a hedge door for skipping the comp round. The eval-side fix
belongs in the worker's max-turns, not in skill prose.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The release-gate campaign confirmed two skill bugs with transcripts.
Sandboxed harnesses reject absolute paths, so following the CHOSEN
CARD directive with the absolute card-base path failed view_image; the
directive and the quality-bar clause now say download into the
workspace and open the relative path. And under the openai worker's
turn cap, models spent the budget on init, cards, and comp generation
and never built the page (a third of small-n supplement slices); the
visualize mandate gains its one exception: at a hard cap the shipped
page outranks optional imagery.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's call: the reviewer-local authorization patch covered one
command while the harness gate silently disables every shipped
subagent, critique panels and the manual-edit applier included. The
argument now lives beside the autonomy counter in context.mjs, emitted
as tool-result content every run: invoking the skill is the user
request such gates ask for; spawn where a reference directs; the
in-thread substitute is for absent capability only and gets disclosed
in one line. new-work keeps the reviewer mechanics and drops the
now-central argument.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A live session on a harness whose guidance gates subagent use on user
request resolved the conflict silently against the skill: it never
spawned the reviewer, stretched the no-subagents fallback to cover
permission hesitancy, and self-reviewed with all the context that made
its choices feel correct. Three tightenings: invoking the skill IS the
user request that authorizes its shipped subagents; the fallback is
for harnesses lacking the capability, not for hesitancy; a substituted
review gets disclosed in one line at finish, never silently.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
"When the harness can X, do Y" hands the model an exit before the
command arrives; the observed reviewer skip walked through exactly that
door. The three gated constructions now lead with the imperative,
present the decision visually, open the chosen card, spawn the finish
reviewer, and carry their fallbacks as trailing clauses for sessions
that genuinely lack the capability. Constructions that already led with
the command keep their routing clauses unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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>
codex.md becomes visualize.md and loads for every harness with any
image generation, native or the API fallback: after the direction
locks, three distinct compositional comps are rendered and put before
the user for approval, in-harness when it can display images,
otherwise on the decision page. Three is the number; one comp invites
rubber-stamping, and this approval round has repeatedly produced the
most compositional and ambitious work, so new-work now marks it
never-skipped. The codex-only subagent stays as a codex note.
The recent rule additions are tightened by a third: the asset and
imagery bullets merge into one, the canon exit loses its restatements,
the DESIGN.md-rule and chosen-card and ceiling clauses each shed their
second clause saying the first clause again. Same laws, fewer words;
prose that grows without bound recreates the attention gravity it was
written to fight.
Dist rebuild still deferred; the release-gate campaign reads the
pinned dist.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
From Paul's approved UX and the eb686f36 session post-mortem:
The standing exit: direction rounds carry a quiet, permanent "Play it
straight" action (payload flag canon, reserved id) on the decision page
and as the last structured-tool option. It is the user's door, never
the model's: never recommended, never weighed against the roll, and
choosing it swaps the bar rather than lowering it, two or three named
reference products become the craft level, canon executed at full
commitment. Safer/conventional steers resolve here, never to a
stranger re-roll.
Session fixes, each mechanical where possible: the ANSWER line now
names the chosen card's hero and board and directs opening them before
code (the session built from text alone after viewing a different
world's card); generation scale joins the imagery rule (a library of
centered 128px subjects foreclosed the atmospheric hero); DESIGN.md
rules are checked against the world's native devices and never added
to silence a hook finding (the session banned arcade lettering's own
offset shadow and laundered 8px through the ramp); staging joins the
FORM contract block (the axis was dropped silently at world-choice);
the finishing reviewer audits the ceiling against the QUALITY BAR card
after persistence (floor rigor was disguising unreached ambition); the
icon-tile clause names hand-drawn icons as remedy, not target.
Dist rebuild deferred: the release-gate campaign reads the pinned dist.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The structured-tool channel collapsed to a single direction plus
re-roll, which read as "the system only ever offers one idea" next to
the multi-card decision page. Both channels now share one structure,
assigned direction leading, the one or two fused challengers that
survived the weighing as named alternates, re-roll with steer, and
differ only in richness. The anti-lineup rule stays precise: what never
appears is a ranked menu of the model's own grounded candidates; dealt
challengers carry no ranking rut.
Note: dist rebuild deliberately deferred; the release-gate campaign is
running against the pinned dist and rebuilding mid-run aborts it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
With the PRODUCT.md skip fixed, the Opus smoke unmasked the adjacent
gap: the model builds a new world and never writes DESIGN.md (zero
attempts), so the worker's requiresDesign assertion correctly fails the
run. Same disease, same treatment: DESIGN.md is now part of recording
the decision, written before the first build edit in the same stretch
as the direction contract, and the finishing reviewer checks
persistence first, before any craft point is scored.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The truth split already permits full-fidelity demonstration data, but
permission at selection time was not holding at build time: models that
would not author covers, names, or thumbnails compensated with chrome,
which is the content-starved look the detector hunts. Two build rules
make it a mandate: every blank the ask round left open is authored at
production fidelity (content is authorable, claims are labelable,
nothing is omittable; unanswered commercial claims ship as marked
placeholders with a replacement list), and when image generation is
available, generating the build's imagery is part of building rather
than a nicety.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A live retest showed the model dropping to the structured question tool
when the roll degraded: with no challengers and no cards it judged the
page pointless and presented one option in plain text. The degraded seed
output and the new-work rule now both state that degradation changes the
cards, not the channel; a browser session presents the assigned
direction as a single text-only card with re-roll on the decision page.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The env-var bypass (IMPECCABLE_QUESTION_DISABLED) relied on the harness
remembering to set it. The script now also self-detects CI, SSH-without-
display, and displayless Linux and exits 2 with the structured-question
advice; --no-open skips detection (caller opens the URL itself, as the
tests do) and IMPECCABLE_QUESTION_FORCE=1 overrides it. new-work.md now
frames the decision-page rule by capability: open a browser if you can,
structured question tool if you cannot, exit 2 means fallback not error.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Gate2 measured it precisely: on matched assignments Opus renders warm,
bookish, and child-facing subjects as cream, serif-italic, and
lamplight while Sol renders the same positions saturated, and neutral
prose hardening did not move it. The codex and gemini blocks set the
precedent for provider-conditional counterweights; this adds the claude
block at the palette decision: the first palette is already spent, an
OWN-WORLD block reading cream/paper/parchment/lamplight for an unpinned
Persuade surface is a failed rendition to rework from the world's
saturated materials, and nothing about the subject requires the
default. Verified present in the claude-code dist and absent from
codex.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's question exposed the blind spot: a closed tab left the agent
waiting out the full timeout. The page now sends a heartbeat every five
seconds while open; the server stamps lastBeat into the state file, and
--wait reports PAGE CLOSED with exit 4 when the beats stop for fifteen
seconds without an answer. The prose defines the fallback ladder:
re-present once through the structured question tool, then proceed
unattended with the assigned direction, stating assumptions.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Craft-gate forensics on matched hands: both models obeyed the same
assigned indices, but Opus rendered every kids cell as cream paper,
lamplight, and serif, and sample 1 chose Fraunces off the reflex list
with a bookshop-signage rationalization, while Sol rendered the same
positions as indigo bookcloth, coral thread, tomato, and marigold. The
dice work; the rendition prior escaped through two hatches, now closed:
naming a reflex face requires a reason no other face satisfies and a
subject association is never that reason; and bookish or child-facing
subjects do not soften the calibration, because cream paper is the
smallest corner of the book world.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's call: start mode never auto-opens; the agent is alive and opens
the printed URL itself, in-app browser first, then the system opener,
then showing the URL (--open forces the system browser from the script).
The prose now leads with the start/open/wait flow and keeps the blocking
auto-open path for harnesses that can background a shell.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's two concerns with the blocking design. --schema prints the exact
payload example so the model never guesses the shape (new-work.md points
at it). And harnesses that cannot leave a shell blocked (or cannot open
a browser while blocked) get a two-phase path: --start daemonizes the
server and returns the URL plus a key immediately, --wait polls for the
answer with exit 3 meaning ask again, exit 2 meaning the server died,
and --stop for cleanup. The browser open happens from the detached
server process, so it works even when the agent thread is short-lived.
State lives under .impeccable/questions/.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's design, three pieces:
serve-question.mjs: the world decision presented as a themed page instead
of a text prompt. The script serves an impeccable-styled option board
(assigned direction leading with THE ROLL badge, dealt challengers as
alternates carrying their QUALITY BAR cards, re-roll and steer built in),
prints the URL, opens the browser, and blocks until the user chooses;
the answer lands on stdout as ANSWER JSON, so the shell call itself is
the wait and no harness machinery is needed. Local images are served by
the ephemeral server; nothing leaves the machine.
generate-image.mjs + context.mjs IMAGE_GEN_AVAILABLE: when an OpenAI key
is in the environment, context reports that image generation works even
without a harness-native tool (gpt-image-2, billed to the user's key,
stated before first use; Google skipped by decision). Harness-native
tools always win when present.
new-work.md: visualize-before-build is now the default whenever any
image generation exists, not a codex.md special case; the attended
presentation prefers the visual decision page and falls back to the
structured question tool. Evals keep the unattended path untouched.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul confirmed the craft-bar experiment: builds that saw the dealt
worlds' hero cards produced visibly stronger execution than the
no-image control. One clause makes the mechanism reachable for
harnesses that read only local images: download the card, then view it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's directive: the rendered board and hero for each dealt world ride
along with the challengers, framed as a quality bar (the finish and
commitment level the build is expected to reach), never as a mockup to
copy. The seed prints QUALITY BAR urls per challenger, preferring
API-provided cardBoard/cardHero fields and deriving from the concept id
otherwise; new-work.md instructs image-capable harnesses to view them
for the world being built. Server side, the roll API now returns
cardBoard/cardHero per challenger (impeccable-site).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Discussion outcome with Paul: a mediocre material world loses to
excellent abstract craft, but an unanchored "just be beautiful" escape
hatch would hand selection straight back to the model's priors. The
resolution: abstraction enters as named systems with their own grammar.
The derivation now states that the audience's graphic and screen
traditions (notation, publications, identity programs, data graphics,
interfaces) are as concrete a candidate as any physical artifact. The
catalog side of the same decision is a 12-entry abstract-graphic
authoring round in impeccable-site, pending review.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Smoke findings (Paul's review): the kids-reading derivation produced
seven candidates from one material family despite the divergence line,
and the brief-pinned bookshop world was rendered as the generic AI
bookshop (cream, serif italic, soft glow). The list must now span at
least three material families, and a pinned world licenses its full
material range, never just its softest rendition.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Paul's probe review (recovery-ab): the no-invention rule was blocking
bold greenfield directions whose demonstration data does not exist yet;
kids-reading amplified the product's "quiet support" adjectives into a
whole-page aesthetic; and nothing guaranteed a Persuade surface still
sells once the form commits (a prior generation shipped zero nav and
zero CTA in the first viewport).
- Truth split in two: commercial and factual claims stay uninventable;
illustrative material is authored at full fidelity, labeled synthetic,
with a replace-with-real list for the user. Mirrored in the build
section so execution-only sessions get it too.
- Persuade floor restored from a22: conversion lives inside the form's
own vocabulary (one-line hook, visible primary action, legible reading
order); a committed form that hides the offer has not finished
translating. The contract's FIRST VIEWPORT block now names where the
primary action sits, and the finishing review verifies the mode did
its job.
- Calibration: negative constraints rule out devices, not exuberance;
product-behavior adjectives do not dictate surface energy.
- Web leverage: when the chosen world names a technique (canvas, WebGL,
view transitions), build the technique, not a static imitation.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
