Waivers follow the context's shape, not the --scope flag: one html page keeps
page semantics under either scope (its claim reaches the sheet it links) and a
lone component beside an unlinked stylesheet answers for its own file only.
Every '<' that starts no tag is a stray spaced out of doc.src (a '<!' or '<?'
with a '>' ahead stays), an unclosed comment runs to the end of the file, the
three bare /<[^>]*>/ strips use stripTags, the copy tier drops open script,
style, template and nav tags nothing closes, component tags are paired once
per document, columnCount counts repeat() arithmetically, an unclosed
@starting-style block is read once, the import scan matches import shapes
only, the render locator refuses a non-letter tag and reads the class
attribute linearly, tickers are answered from the tree, a tooltip duration is
not a delay, class tokens are appended in a loop, the emoji dedupe is keyed by
the icon, the emoji classes spell U+FE0F and U+200D as escapes, the contrast
and focus viewports are named once, and the hero-fold title matches what the
probe measures.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
test/design-lint-seventh-cycle.test.ts (13 tests) pins the stray-tag tail,
the shadow-layer split, the JSX generic and ancestor-closing cases, the
one-mode token table, the per-file component claims under both scopes with
their own suspected rules, the stylesheet beside components, the Reporter
file key, outline: none, the sign-in form, the shared tracking ceiling and
the missing-stylesheet warning. docs/design-lint.md: the token block and its
mode overrides, the stray-tag note, the component-directory claim scope,
the email-band exclusions, the missing-sheet warning and zero-chroma-neutrals
as a fourth narrow reading.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
A tag start the scanner finds no '>' for is spaced out of doc.src, so every
'[^>]*>' scan over a page (the lint's and the craft check's) costs the page
its length; render's line lookup reads the same text. The JSX scanner skips
a generic element's type arguments and closes every element under a close
tag that names an open ancestor, as a browser would; a closer naming nothing
open is text. The token table is read in one colour mode: the base block's
last declaration wins and a prefers-color-scheme or data-theme override only
fills tokens the base never declares; a custom property declared in any
scope is a token definition, not a literal; a token finding points at the
declaration that won. In a directory of components, under either scope,
each file's claims answer for that file alone with its own suspected rules.
The Tailwind dedupe keys files the way Reporter prints them. outline: none is
not a ring for input-state-shift; a form with a password or a sign-in button
is not an email band; one CONTROL_TRACK_MAX_EM (0.04) for both halves of
wide-tracking; one CONTRAST_SLACK; hairline-under-shadow splits layers with
splitTop; a missing linked stylesheet is warned about; TwItem.toks and the
controlHeight scope guard are gone; zero-chroma-neutrals is titled for what
it reads. aside-render sets the device override or viewport only when a
sized step's size changes and clears it before an un-sized step.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
test/design-lint-sixth-cycle.test.ts (18 tests; 13 red at 3f0c62582) pins
the per-page work, the right-place findings, the rule judgements and the
waiver and render cells the state matrix named. The emoji-tile cap is
asserted exactly, the pixel-pass test says what it checks, the recorded slop
page now names its generic hero copy. docs/design-lint.md: the real-path
bound, the separator note, page and component claim scopes, hero-fold's
measured half, two more items the lint does not cover, the heading-pattern
exemption and the digit-range dash.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
line-length computes its floor only for a measure between 28 and 45ch and
answers shortProse once per class; wide-tracking resolves inherited case
through a prefix index instead of a scan of every rule; the JSX scanner
fails every ancestor of an element that ran to the end of the file at once;
the hairline-under-shadow pair reads class tokens instead of a two-sided
regex. tooltip-focus-delay names the sheet that holds the :focus delay;
the placeholder emoji scan reads the comment-blanked body for every kind;
fake-send finds its line in a component's script text. A stylesheet href is
bounded by its real path too. The LINT line's separator is kept out of
page-supplied text. Waivers: a component claim answers for its own file,
weighed against its own text; the Waiver shape is a reason string; refusals
go through Reporter. Render: a later fail replaces a warn with the same
evidence, a non-string fix is dropped, the timeout is a constant. Rules:
a meta label in mono is a costume, a leading-dot length reads as a fraction,
a percentage leading collides like a unitless one, an italic word after an
icon is still read, an italic rule on the word is not a display system, en
dashes between digits are ranges, and the catalog's two hero lines are
generic-hero-copy. The Tailwind parse cache and its eviction are gone.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
test/design-lint-fifth-cycle.test.ts (21 tests) pins the comment-dead placeholder, the linear-time digit run, 699k-rule sheet, unclosed JSX and tracked-label pages, per-sheet waivers and the remaining state-matrix cells, every wrong-shaped render field, and the rule judgements; the fourth-cycle shell test asserts the observable effect, the RSS assertion is structural, time bounds are widened for shared CI shards, the stale empty block is gone, and the FREE_FIXTURES inventory names both cycle files. docs/design-lint.md states the project-root bound on relative hrefs and the three narrow readings (whole-page focus ring, unconditional 4px scale, scroll-container-blind double-sticky).
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
The placeholder emoji scan reads the comment-blanked text, so markup inside <!-- --> is dead for every rule. A digit run, a long mono word and a run of scheme prefixes cost their length (FACT and UNSPLASH_URL only start at a token, twDataText bounds its text); a 2 MB sheet of 699k rules no longer blows the stack (rules are appended in loops); JsxScanner spends one failure budget across jsScan and el; the Tailwind had() lookup is a Set; accentedHeads indexes rules by class once; contexts drop their caches once their families have run; the parsed-token cache is bounded. treeOf gives a close tag its parent so a fact after an inline close is counted in its section. Waivers are keyed per rule for a page claim and per rule@sheet for a stylesheet claim, so two linked sheets each answer for their own findings. The render tier validates every page-supplied field (at.cls, ring colour and ratio, probe fg and threshold, PAGE_ERRORS entries, a malformed entry naming the render rule) and keeps its measured findings; lineOf answers from one line index. Rule judgement: a textarea's min-height is not control-height, attribute-selected inputs are read for input-state-shift, a glyph-only <i> is not an italic word, a bounce keyframe by name is bounce-easing, single-quoted inline styles are read, a project root is a package.json or .git. Housekeeping: Probes typed from PROBE_FNS, EYEBROW_* shared from chrome-eyebrow, WaiverClaim.raw and dead imports removed, evidence wording for tracked-only controls and black glass, emoji icon tiles capped like every other emoji path, unreadable files get their own fix text.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
heroClasses reads the hero through heroIndexOf, check64 uses twMonoIsMaterial, check67Html imports capsRatio from chrome-eyebrow, sideStripes loses its bare block; a settle test drives confirmPixels with the swatch fixture (a claim dropped by the pixels behind it, one rewritten, a weak ring confirmed).
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
test/design-lint-fourth-cycle.test.ts pins commented-out markup dead for every rule, unclosed <li> Tailwind frames and wireframe slots, linear-time unclosed runs and component scripts, cyclic var() and nested @media, leading-dot lengths, a symlinked lint root, oversize files reported, a whitespace-named stylesheet's waiver, the overused-font stems, width: min() caps, box-shadow rings, Meet headings below the hero, wrong-shaped render answers and the never-waivable refusals by class; the settle golden no longer repeats the kept width and expects the bad grade as tampering. docs/design-lint.md describes the caps, implicit closes, skipped-file findings and the matching rules.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Commented-out markup is dead for every rule (Doc.src is the comment-blanked text; craft copy, the module-script check and the border+shadow scan read it). Headings, buttons, links, labels, selects, forms and an open <p> close the way the browser closes them, a tree nests at most 512 deep, and the heading, footer-column, sign-up-button and span readers walk the shared pairing instead of lazy regexes, with each element's text computed once (textOf). Tailwind frames and wireframe slots answer from treeOf. resolveVars stops on a cycle and caps its size; every parseCss at-rule branch honours the depth guard and an inline <style> that fails to parse is skipped with a warning. cwd and lint roots are read through their real paths, a skipped source file is one internal-error finding, relative refs stay inside the project, a linked stylesheet over 2 MB is skipped. overused-font and banned-font match a catalog stem to a family name's first words; width: min()/clamp() caps a container; a box-shadow focus ring beside a border-colour change is not input-state-shift; Meet and triad headings below the hero are suspected; the stylesheet-waiver key goes through oneLine; pxOf reads leading-dot lengths; check67 precomputes selector bases; the render tier reports a wrong-shaped answer per probe and keeps every measured finding, flags a bad grade as tampering, and never repeats the kept width when merging horizontal-scroll.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Waivers settle once per page after both tiers, so a blocking rule only the
render tier suspects is waivable and one it measures is refused; a claim in a
stylesheet answers for that stylesheet on every page that links it (no
shared-word check), so a sheet shared by two pages is no longer waived on one
and refused on the other. Every printed field is one line with control
characters stripped; the waiver regex is linear and ends at a carriage
return; a root-relative href cannot climb out of the lint root. The ancestor
map closes optional end tags implicitly and pops to the matching open tag, so
unclosed <li> no longer makes it quadratic. The static tier reports outline:
none with no ring anywhere as missing-focus-visible at its line;
literal-outside-tokens needs a --color-* or --font-* token to exist; the
banned gradient includes indigo-to-violet; a new banned-font rule reads
gstack's BANNED_FONTS; MONO_FACE matches whole words; bold keywords and
tokens count as weight 700 for the contrast threshold. stylesFor reads only
the entries that can match; glOcclusion memoises the pinned walk; the
confirm screenshot is taken at 1x; a page that throws in the render tier is a
render finding and the CLI exits 3 on an internal error. Docs name the token
conventions and what the lint does not cover.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Waivers reach render findings; a blocking rule the lint only suspected is
waivable like any ask, a measured or absent Floor stays refused. A stylesheet
a page links from another directory belongs to that page and is not linted
again on its own; a stylesheet linted alone keeps its waivers. Root-relative
hrefs resolve against the lint root. Stylesheets are read and parsed once per
run. A rule family that throws reports internal-error at ask instead of a
silent pass; aria-labelledby and probe tag names are escaped before they
reach a RegExp; a probe result naming an unknown rule is a render finding,
not a crash; fix strings are one line. overused-font follows gstack's
catalog of display faces; one MONO_FACE list for every monospace check.
parseCss caps nesting depth; medianRgb uses histograms; glHide batches its
layout reads. Dead exports removed (isPage, openTags, over, ComponentTag,
PixelRect, the ids re-exports).
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Ports hallmark's sloplint-render.mjs onto lib/aside-render: one tab per
page, the in-page probes installed once and run at each viewport, a second
pass with full-page screenshots only when a contrast claim needs the painted
pixels (a small PNG reader replaces canvas sampling). Gates 23, 25, 26a, 34,
40, 41, 44a-d, 49a/b, 58, 71, 72, 73, 75, 77-83, F17-F20, plus the page that
did not render. The stamp's posture, hero-shape, photo-shape and genre
guards are gone: the strict form applies. Focus rings are read by focusing
each control in turn (programmatic focus matches :focus-visible on the
daemon with no prior pointer input; a stylesheet scan is the fallback).
lib/aside-render: an eval step takes width/height like a screenshot step
(device metrics on Aside, viewport on browse, cleared after), and spec.file
may be an http(s) URL, opened as it is with nothing served.
The clean fixture gained border-box on its hero, a centred, taller hero
visual, a wrapping actions row, a statement-sized close head and a shorter
story paragraph: the browser showed the action below the 1280x800 fold, a
dead band under the hero, a 600px image spilling past a phone screen and a
thin close section.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Chrome (gates 19 42 46a 47 54 63 68 69 70 85): gate 19 split by tell onto
aphoristic-cadence, generic-hero-copy, generic-cta-copy, em-dash-overuse and
template-footer-line in rules/chrome-copy.ts (with the two-filled-buttons half
of F4 and the shared prose mask); the eyebrow in rules/chrome-eyebrow.ts; the
nav and footer shapes, invented proof, drawn chrome, placeholder services,
template punctuation, wireframe slots, repeated facts, the fake send and a
static version of the email-band close in rules/chrome.ts.
Assets (gates 30 33 40 41 49 76): emoji icons, mixed icon libraries, unlabelled
svg/canvas, contrast pairs and surface inks (suspected at ask in the static
tier), wrapping labels split into button-label-wraps and link-wraps, broken
images and stock hotlinks.
Finish (F1-F9 F15 F16 F21-F24): all polish except fake-avatar.
Tests cover every id with a negative control and a planted positive; the clean
fixture gains scrollbar-color so its committed palette themes every browser
surface.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Replaces the tailwind.ts stub with sloplint's utility reader (twParse,
twInfo, twGroup, variants, stops, sizes, tracking, families, shared
project tokens) and the tells it feeds, and fills rules/component.ts
with checkTailwind and checkComponentText: layout-transition,
bounce-easing, purple-blue-gradient, gradient-text, glow-wash,
literal-outside-tokens, hover-tells, wide-tracking, kicker-above-heading,
monospace-costume, framework-palette, vh-height, template-cards,
template-punctuation and emoji-decoration over class strings in HTML,
JSX, Vue, Svelte and Astro files and cn()/clsx()/cva() calls.
test/design-lint-tailwind.test.ts holds a negative control and a planted
tell per rule plus the tailwind-fail, tailwind-pass and clean fixtures.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
rules/color.ts: gates 2a 2b 7a 7b 22 23 29 48 65 74 (check2, check7, check22,
check23, check29Glow, check48, check65, check65Tailwind). rules/type.ts: gates
1 25 37 38i 38ii 38iii 55a 55b 64 67 (check1, check25, check37, check38a,
check38c, check55, check64, check67, check67Html). Thresholds and heuristics
are hallmark's; evidence and fixes are rewritten in gstack's voice; the stamp,
genre and posture guards are gone, so every check runs in its strict form.
The clean fixture carried two real tells hallmark's own sloplint also flags:
an oklch literal on .button:hover (48) and the cream paper + Fraunces + clay
accent look (65). The hover colour is now a token and the display face is
Literata.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Recreates Dependabot's #3307 on a crew branch so it can take a version stamp
at merge. 6f2ac81f5cd81fe351a197ba0b00d91d03cd2aee is the commit the
annotated tag v4.39.0 points to.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(gstack-lease): take the lease lock with mkdir, not ln -s, so Git Bash on Windows stops waiting out 300 tries (#3330)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* test(gstack-lease): count lock retries instead of timing them
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(lease): a dead lock that cannot be renamed aside no longer spins lock()
break_lock ignored drop()'s failure and returned 0, so lock() continued past
the tries limit forever when a dead holder's .lock could not be renamed aside
(a Windows file in use). It now removes the lock in place under .lock.break,
and returns 1 when that also fails so the try is counted. touch no longer calls
a dead holder 'live' in its exit-3 message.
Adds the /ship coverage-audit tests for the lock's failure paths.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(lease): pre-landing review fixes
- drop() reports a lock released once it is renamed aside, so a failed
delete of the renamed copy never makes unlock or break_lock rm -rf a new
taker's lock at .lock.
- An ownerless lock is broken only once it is 2 s old, so a live taker whose
claim landed after a breaker read the claims keeps the lock it is about to
publish. An empty owner, an invalid pid or a stray file counts as ownerless.
- touch's prune stays out of the lock (top-level lease files only) and clears
leftover renamed-aside locks; a failed claim write leaves no claim.
- Docs and CHANGELOG state the mixed-version and dead-break-lock limits
instead of claiming exclusion that does not hold.
Tests pin each fix (each fails against the previous bin), make the exclusion
case wait on a release file, prove the parallel case is mutually exclusive,
and check the Git Bash holder keeps its lock.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(lease): second /ship review pass
- orphaned() reads the clock before the lock's mtime and requires the mtime
unchanged at the end, so a lock made meanwhile is never removed; an mtime
a minute or more ahead (clock skew) counts as old.
- Claims older than 2 minutes are abandoned and pruned by touch.
- A waiter skips the claim and mkdir while the lock is visibly held; the
wait is bounded at 300 tries or 30 s; a non-numeric tries value falls back
to 300; ten quick retries (not three) before a lock nobody holds is
reported as untakeable; pid 1 is a valid holder again.
- Tests: deterministic young-lock clock, skew and non-numeric owner shapes,
abandoned claims, the deadline, explicit 30 s timeouts.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(lease): third /ship review pass
- A claim is honored while its claimant lives; only a claim with no start
identity expires after 2 minutes, so a stalled live taker keeps its lock on
hosts that record identities (the remaining case is a documented limit).
- A late taker never overwrites an owner already in its directory, and does
not remove a directory that has one.
- A lock whose owner file exists but cannot be read is never ownerless.
- Zero-padded owner pids count as dead; --help prints the whole usage;
touch prunes only claims whose claimant is gone or recorded no identity.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* docs(lease): scope 'never by age' to locks that name a holder
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
---------
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(windows): shell bins pass NUL as Bun's empty config when MSYS path conversion is off (#3415)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(windows): detect Git Bash/MSYS2/Cygwin from OSTYPE, as gstack-doctor does, so uname stubs in setup fixtures stay untouched
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix: pre-landing review fixes (#3415)
- Choose NUL only when cygpath is also on PATH, so an OSTYPE=msys inherited
by a POSIX bash keeps /dev/null instead of a cwd-relative NUL file.
- gstack-developer-profile --log-session reports a Bun failure instead of
skipping it as invalid input, like gstack-timeline-log.
- learnings-injection tests check that their anchored bun -e block exists.
- CHANGELOG names the callers this fix does not cover; curation reason fixed.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix: pre-landing review fixes, cycle 2 (#3415)
- timeline-log and developer-profile validators print their verdict; any
non-zero Bun exit is a reported failure (a Bun crash on Windows exits 3).
- timeline-log keeps the original event when ts stamping fails.
- The definition ratchet scopes a local GSTACK_BUN_CONFIG to its function.
- The shell-bins fixture clears its preload marker per test.
- CHANGELOG states the callers that discard the report and the Bun 1.3.x case.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* docs: CHANGELOG notes Bun 1.3.10 evidence and remaining literal /dev/null callers (#3415)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
---------
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* feat(review): /review re-reviews only its fixes after an edited pass, with a final full pass (#3339)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(review): fix-range narrows only DIFF_BASE outside Step 4.8; Step 5.0 ignores this invocation's records (#3339)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* test(review): fix-range walk logs each pass's Step 4.8 attempt in /review's real order (#3339)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
---------
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* feat(design-html): craft section and free craft check, adapted from Elaya design skills (GSTA-92)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(GSTA-92): restore the full CHANGELOG under the Unreleased entry; drop a $-digit literal from craft.md; count only craft-check executions in design-html-slop-gate
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
---------
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(context): record the save commit and flag work that landed after it
/context-save recorded the branch, a timestamp and the modified files, but not
the commit, so /context-restore could not tell that work landed after the save.
A checkpoint saved before a step was committed came back with that step still
listed as the plan.
- gstack-slug --stamp-checkpoint also writes head:, the commit the save
describes, next to remote:/project_root:/worktree:.
- context-restore compares head: with HEAD before presenting. When commits
landed after the save it prints how many and lists them, and moves Remaining
Work under Verify first; a saved commit that left the branch history is
reported the same way. Checkpoints saved before this keep their old sorting
with a one-line note.
- test/context-restore-head.test.ts runs the real save-path, stamp and
head-check blocks in fixture repos: unchanged, moved, diverged after a reset,
pre-tracking, and a malformed head: that must never execute.
* fix(context-restore): name another branch as a cause when the saved commit is not in history
Cross-branch resume (Conductor handoff) is a supported restore path, and there
the checkpoint's commit is usually not an ancestor of HEAD. The diverged line
said only 'rebased, reset, or not fetched'. It now names 'saved on another
branch' first; the new test case covers a save on a feature branch restored
from main.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* docs(changelog): unreleased entry for the context head check (port of #3316)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* fix(context-restore): head check passes the fence lint, reads frontmatter only, prefers worktree:, and overrides provenance sorting
- The block used a free-text placeholder, <the chosen file>, which the
generated-bash lint (test/env-host-fence-conformance.test.ts) refuses on
every host. It now uses the allowlisted <the FILE path printed above>, with
prose naming the Step 2 choice.
- head:, worktree: and project_root: are read from the leading frontmatter
only (sed, no positional $N tokens Claude Code would rewrite), so a head:
line in a checkpoint's notes no longer counts.
- Outside git, the comparison uses the saved worktree: before project_root:,
so a save from a nested linked worktree is not compared with the main tree.
- The provenance rules now say they are skipped after HEAD_MOVED,
HEAD_DIVERGED or HEAD_UNKNOWN, instead of contradicting the head check.
- The re-stamp test checks the exit status and that the head moves.
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
---------
Co-authored-by: Yao Amevi A. Sossou <47247401+soyames@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* test: flaky-fix pass — fragment-match reads the agent's report, ff-update reap counts its own sleepers, install-admission gets a measured timeout, judge parse errors name the judge (GSTA-91)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* test(qa-only): late scenarios freeze the fixture clock during setup and register their real outer budget (GSTA-91)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* test(harness): record the llm-judge parse-message change as non-behavioral (GSTA-91)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
---------
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
