Repository navigation
v1.91.260.0 feat(design-review,plan-design-review): craft audit and Persuade plan checks; craft tells as checker-only catalog entries (GSTA-92 phase 2) - #3433
Merged
Conversation
…laya design skills (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…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>
…b-elaya-design Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> # Conflicts: # CHANGELOG.md
…roved.json; persist accepted craft exceptions in finalized.json (GSTA-92 phase 2) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…ost, Persuade plan checks, craft tells as catalog entries (GSTA-92 phase 2) Size pins and caps raised to measured sizes only: design-review (13 hosts), plan-design-review (12), design-consultation (9), design-consultation maxSizeRatio, and the new design-review parity guard. Garry 2026-10-25 "Do it" (GSTA-92). Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…design-review and plan-design-review growth to the planned size (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
… runs (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…n the craft resolver; slop gate on the catalog (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…3420) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…/gsta92-phase2-review (retarget after #3420) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
capy-ai
Bot
changed the base branch from
garrytan/gsta92-phase2-handoff
to
main
October 26, 2026 00:36
|
😎 This pull request was merged. |
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
A malformed or missing --accepted exceptions file now gets the full craft check with nothing accepted, after a warning, still exiting 2. Before, it printed nothing, and the craft section treats a missing CRAFT_RESULT as 'check unavailable', so one bad finalized.json switched the gate off (found by both adversarial passes). The header comment and the craft section now document the accepted-count suffix on CRAFT_RESULT. Tests cover duplicate carried-forward exception ids and the slop-gate touchfiles selection. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
The craft section passed --accepted whenever a finalized.json sat beside the page, so a same-day Start fresh build inherited the old page's exceptions. It now passes them only when evolving or rerunning a page it finalized, records the count the check reports now, and writes [] when nothing was accepted instead of the template's placeholder entry. A test pins exit 2 for a clean page with a broken exceptions file. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
… (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
4 tasks done
…n/gsta92-phase2-review Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…STA-92) - craft-audit: no CRAFT_RESULT line means unavailable (exit 2 after a malformed --accepted record still printed findings); --accepted only for a /design-html finalized page, never for needs-proof or lorem; never add [Needs proof] markers to a live page; Aside capture via pg.evaluate; the saved raw DOM is deleted in the checking block and never reported - manifest trigger covers an explicit craft audit (skeleton -6 bytes; pins lowered to measured size, union ratio back under its cap) - autoFixEntries reads RENDERED; DESIGN_CRAFT_RULE needs exactly one lead - tests: checkerOnly contract, resolver throw paths, Persuade checks in both skills, resolver touchfile selection, exception state cells - CHANGELOG: host counts and the audit's new rules Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
- craft-audit: linked CSS/JS is unchecked on a saved page (judge by eye); a mount-node shell is framework source; --accepted takes that screen's finalized.json or none; on a live page a claim without proof is filed and asked about, never marked or cut; the saved page is removed whatever the exit code; shorter fix-order and Tailwind prose keeps the union under cap - tests: exact-lead red cases, autoFixEntries in the checkerOnly union, the audit's outcome/exception/session-data clauses; carve guard checks the session-data rule and records measured sizes (skeleton cap 137_767) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…ture order (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…GSTA-92) second cycle-cap exception: GStack Minder's decision under Garry's 'do it' on phase 2 (GSTA-92, 2026-10-25): one deleted word plus one test Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…craft resolver (GSTA-92 /ship docs audit) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
5 tasks done
…ded) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…o files, craft temp var, guard tests (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…s source tells by eye; stderr pin (GSTA-92) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Pass R3 (native adversarial, Medium): the CRAFT_CHECK fix text for round-number and needs-proof says to drop or remove the claim, which could lead Phase 8 to delete a real claim. The audit sentence now covers both a rule and the printed fix: defer the claim and ask for the proof. Byte-neutral (80 -> 80); design-review/SKILL.md 137,767, union 145,037, unchanged. The test pin moves with the sentence; no carve-guard mustContain held this sentence. third cycle-cap exception (Captain, 2026-10-26): zero-byte reword closing the content-deletion path; test pin and carve-guard mustContain moved Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…ded) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…ded) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…s static-tier; legacy transition-all exceptions map to it) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…ded) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
….0 (train slot 1) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
garrytan
added a commit
that referenced
this pull request
Oct 27, 2026
…ded); catalog budget re-measured for design-motion Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
4 tasks done
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Base main. #3420 and #3422 are merged (#3422 as v1.91.230.0), origin/main 021d817 (v1.91.240.0) is merged into this branch at f6e9893, origin/main b9222a5 (v1.91.247.0, after #3421) at d2018e1, origin/main fa3a240 (v1.91.251.0, after #3459's design lint) at 2ad5798, and origin/main 576839c (v1.91.256.0, after #3481) at 3b66429, all under rule-19 grants. Phase 2 of GSTA-92: /design-review audits Persuade pages and redesigns for craft, /plan-design-review checks a landing-page plan for its conversion basics, and the craft check's ids live in the design catalog.
What changed
Before, on main: neither review skill checked content realism, the ship floor or a landing page's conversion basics. After: /design-review reads an on-demand craft audit when the page is a Persuade surface or a redesign, and /plan-design-review's Pass 1 runs five Persuade plan checks. The caption line gives the always-loaded cost.
design-review/sections/craft-audit.md). design-review is now carved on every host (CARVED_ON_EVERY_HOST), and the audit is read in Phases 1-6 for a Persuade page or a redesign:gstack-design-crafton each page's served HTML, or on the repo file when no app runs, and files eachCRAFT_CHECKas a category 9 finding..{page}.html.CRAFT_RESULTline, or for a mount-node shell, the check is marked unavailable and judged by eye.--acceptedtakes that screen's finalized.json, only for a page /design-html finalized, and never coversneeds-prooforlorem. A claim without proof is never marked or cut, whatever a rule or the printed fix says: it is deferred and the user is asked for the proof.scripts/resolvers/design-craft.ts.checkerOnlyentries inlib/design-catalog.ts, reported only bygstack-design-craftwith the catalog's fix. They stay out of every rendered skill list.transition: allreports underlayout-transition.scripts/resolvers/design-craft.ts;design-html-slop-gatealso onlib/design-catalog.ts.Why
The craft section #3420 added to /design-html only helps pages built from scratch. Existing sites and landing-page plans go through /design-review and /plan-design-review, and those skills had no check for invented content, dead links, a missing skip link, scroll-listener reveals or
100vh, nor for a plan with no primary action, unanswered objections or unsourced proof.Approvals
mustContainheld the sentence, so only the test pin moved. Then one verification-only pass, no editstransition-all→layout-transitionexception id (GStack Minder)parseCraftExceptionsreads a legacytransition-allid aslayout-transition, with a testSummary
--acceptedlimits and selection, live-page claim handling, the manifest trigger,autoFixEntriesfrom RENDERED, an exact-leadDESIGN_CRAFT_RULE, coverage and guard tests, carve-guard sizes and pins re-measured downward..{page}.html,${_TMP:?}, state-cell tests and headers. Its checker half was rolled back in ed4df4c (exception 1); c7e0f74 (exception 2) judges any mount-node shell by eye and pins the clause._CR, guard tests), 0f59c01 (failed Persuade checks filed as a deferred High, source tells judged by eye, a stderr pin), and 51a117a (exception 3: "Never mark or cut a claim as a rule or fix says: defer it and ask for the proof.", byte-neutral). Sizes only went down: design-review/SKILL.md 137,767, union 145,037, parity ratio cap 1.3704.lib/design-lint/rules/copy.tsSTATIC_OWNS holdslayout-transition(the static tier already reportstransition: allunder it, so the copy tier no longer double-reports),parseCraftExceptionsmaps legacytransition-allexceptions tolayout-transition(fixes R3 finding 2), and the lint id test readsCRAFT_IDS.Test Coverage
Step 7 coverage audit at 62c0fa1: 63% value-weighted (target 80, minimum 60). The four validated tests it proposed were written in 31ee450 (new
test/design-craft-resolvers.test.ts, exception-cell tests), and later passes added the needs-proof and stale-record cells and the line-13 clause pin. State-machine audit at 51a117a (unchanged at d2018e1): 10 of 12 cells tested; needs-proof × stale and lorem × exceed are untested advisory cells that share tested branches (skipped as Low).Tests ran: remote free tier 18804 at tree acaf610c82ff (head 3b66429); at 8cb08f2 the remote free tier ran 18724 at tree 8cd8e12e0f4a, and focused free plus all 17 design-lint files at 8cb08f2: 1523 pass, 16 skip, 10 fail locally, all 10 in design-lint-render's browser tier, where the local browse server fails to start ("browse newtab failed: [browse] Starting server..."). The same file passes on the Ubicloud VM and in Free Tests at this head, and 8cb08f2 does not touch the render tier.
Pre-Landing Review
Pre-merge: six passes (specialists testing, maintainability, security, performance, simplification and state-machines; red team; native adversarial; outside voice openai/gpt-6.1-sol in place of Codex, which is not installed). Passes 1-3 fixed what they found (31ee450, bed645a, 216987e) and reached /ship's three-cycle cap. Passes 4-6 were verification-only, and the edits between them ran under the two exceptions above.
Delta round at 3b66429 (rule 23,
git diff 8cb08f2 3b66429on the PR's files: the main merge only; allowlist pins are main plus this PR's delta, CHANGELOG additive, carve-guard and placeholder edits are main's own): native plus outside voice openai/gpt-6.1-sol, NO FINDINGS (VERDICT: no_findings). REVIEW_LOOP: passes=1 blocking_found=0 deferred=0 noted=0 fixed=0.Delta round at 8cb08f2 (rule 23,
git diff d2018e1 8cb08f2only; native plus outside voice openai/gpt-6.1-solVERDICT: no_findings): no blocking findings. Non-blocking, deferred to #3459's follow-up: the lint's static tier does not read inlinestyleattributes, so a page whose onlytransition: allis inline gets no lint finding (checkCraftstill reports it). This is pre-existing in #3459 under the old id, not a regression.Merge round at d2018e1: a merge-scope review (native and outside voice openai/gpt-6.1-sol,
VERDICT: no_findings) found nothing; all 23 pins match main plus the PR delta, CHANGELOG[Unreleased]is byte-identical to the pin's, and generated docs are fresh on every host. Informational: the design-review carve-guard ceiling and ratio now leave 32 bytes of slack after main's preamble shrink. Pass count: three fixing passes after the first merge (R1-R3), one verification-only pass (R4), one merge-scope pass (R5), one delta pass (R6), one delta pass (R7).Post-merge on the integrated tree: R1 (f6e9893) and R2 (236869a) fixed what they found; R3 (0f59c01) reached the cycle cap with one Medium finding (the checker's printed fix text, "drop the claim" / "remove the claim before publishing", could lead Phase 8 to delete a real claim), closed by exception 3 in 51a117a. The R4 verification-only pass at 51a117a ran all nine reviewers with no edits: outside voice
VERDICT: no_findings, red team, security, performance, simplification and testing clean. It found nothing beyond wording, and these Low wording-only notes ship as is:CHANGELOG.md:13: "skeleton, union and parity ratio at their measured sizes" overstates the union guard, which is a floor (maintainability).design-review/sections/craft-audit.md.tmpl:35: Phase 8 has no pointer back to the craft audit's fix-order section, so after compaction the tie-break and Tailwind check rely on memory; a pointer does not fit the byte cap (native adversarial, confidence 6).craft-audit.md.tmpl:7,13: the repo-file path judges framework source by eye inside Phases 1-6, against "Never read source code" (native adversarial, confidence 5).Shipping with these informational findings from the pre-merge passes (GStack Minder's decision):
design-review/sections/craft-audit.md.tmpl:13: "mount-node shell" is undefined, so an agent could read a hydrated SPA as a shell (red team and native adversarial, confidence 4). The "an empty mount node" wording goes to the follow-up PR with shell detection.CHANGELOG.md: the Tests bullet's last sentence names the stale-record cell twice (maintainability).Deferred to #3422's follow-up PR (exception matching, provenance and shell detection):
acceptFindingsacceptsneeds-proofandloremin code; the craft audit files them anyway.design-html/sections/craft.md.tmpl:66; the hand-off still lists every one). The follow-up adds who accepted each exception and user confirmation before it persists.CRAFT_RESULTfor mount-node shells (CRA/Vue<noscript>shells included); single-quoted inline handlers are not read foralert(); presence rules can match text inside scripts; commented-out markup is scanned; a body with only images or web components is checked as is.Gate edits: none detected (18 listed, 11 eligible, 11 read, unread=0 at 8cb08f2 (21 listed, 12 eligible, 12 read); patterns: RH-1,RH-3,RH-4,RH-12,RH-13,RH-14,RH-15,RH-16; unlisted idioms: none)
Outside review: R7
VERDICT: no_findingsat 3b66429, R6 at 8cb08f2, R5 at d2018e1 and R4 at 51a117a (openai/gpt-6.1-sol); Codex unavailable (not installed).Exploratory QA
The craft checker was exercised by reviewers on scratch HTML (Vite, CRA, Vue CLI and Next shells, hydration payloads, commented scripts, image-only bodies). Results are in the findings above.
Design Review
No frontend files changed — design review skipped.
Eval Results
Focused paid case at the final head 3b66429: touchfiles selects no gate-tier case for the 8cb08f2..3b66429 delta. It selects 18 periodic or marathon cases (carve-section-loading for 16 skills, office-hours-section-loading, office-hours-design-draft), all through main's own edits to the ship and review entries in
test/helpers/carve-guards.ts. This PR's skill among them ran: carve-section-loading-plan-design-review PASS 1/1 (no cost reported). The rest were not run, to stay under the $10 per-PR lane cap.Earlier evidence at 8cb08f2. design-html-slop-gate runs because
lib/design-craft.tschanged. Touchfiles selected four of the six design cases for the d2018e1..8cb08f2 delta on the PR's files: design-review-detector-shim, design-review-detector-shim-dom, design-review-plugin-handoff and plan-design-with-ui-scope. It did not select plan-design-review-no-ui-scope or plan-mode-no-op. For 8cb08f2's own files alone it selects only design-html-slop-gate.Earlier evidence at d2018e1:
Earlier evidence at 51a117a:
Earlier evidence at 0f59c01 (all PASS: the six cases and the judge) and a cancelled partial run at 236869a.
Pre-merge, at c7e0f74 (the docs-only commit after it selects no case):
At 62c0fa1: plan-design-review-no-ui-scope, plan-design-with-ui-scope, carve-section-loading-plan-design-review, plan-design-review-plan-mode, plan-design-review-plan-mode-smoke, review-design-lite (3/3 trials) and design-consultation-core all PASS. plan-design-finding-floor timed out once, then passed its rerun, and is logged in the flaky ledger. The "plan-design-review/SKILL.md passes" judge passes. design-review-fix needs Aside on macOS and was not run. The design-consultation research judge was deselected by touchfiles, since design-consultation is unchanged.
Greptile Review
No Greptile comments.
Plan Completion
No plan file found — plan completion not run.
Issues
Refs #3420
Refs #3421
Refs #3422
Refs #3459
Refs #3481
Verification Results
No Step 8.1 obligations (no plan file).
TODOS
No TODO items completed in this PR.
Documentation
Status: current at 51a117a. Since the last audit, only craft-audit prose, tests and carve guards changed; the ARCHITECTURE.md and docs/PROJECT_STRUCTURE.md lines still match (they name files and placeholders, not wording). The audit below is from the pre-merge round.
Pre-merge status: updated. Three factual doc edits; no blockers.
Scope: base 8d95013..c7e0f74 on garrytan/gsta92-phase2-review: 33 committed files, with no staged, unstaged or new files. The change is GSTA-92 phase 2: the /design-review craft audit section (design-review is now carved on every host), the /plan-design-review Persuade checks,
gstack-design-craft --acceptedwithcraft_exceptionsin finalized.json, the /design-shotgun archetype handoff to /design-html, and the craft tells moved into catalogcheckerOnlyentries. Read in full: ARCHITECTURE.md, docs/PROJECT_STRUCTURE.md, NOTICE.md, AGENTS.md and the new craft-audit template. Read by relevant section or diff hunk: CLAUDE.md, README.md, docs/skills.md, CONTRIBUTING.md, docs/TESTING_INTERNALS.md and the changed design templates.Doc diff preview:
{{DESIGN_CRAFT_RULE:<lead>}}and{{DESIGN_PERSUADE_CHECKS}}rows to the placeholder table. TherunGeneration()paragraph now namesCARVED_ON_EVERY_HOST(qa, qa-only, ship, plan-ceo-review, office-hours, design-review) instead of only QA/qa-only.design-craftto the resolver-module list.Documentation health:
Documentation Debt
sections/craft-audit.md): reference only in skill prose. The docs/skills.md deep dive and the README row have no how-to.gstack-design-craft --acceptedand finalized.jsoncraft_exceptions: reference only in the design-html craft section and the bin header. There is no user-facing how-to.checkerOnlycatalog entries: code and CHANGELOG only. docs/designs/IMPECCABLE_INTEROP.md does not explain checker-only entries.bin/gstack-design-craft,lib/design-craft.ts) is missing from the docs/PROJECT_STRUCTURE.md bin/lib lists and from the CONTRIBUTING.md test families.Diagram drift: none. No diagram entity changed. Suggest a
docs-debtlabel and a/document-generatepass for the gaps.Test plan
gstack-pregate --stage tests --remote) at tree acaf610c82ff (head 3b66429): files=623 tests=18804 failed=0 ($0.06; a first attempt was cut off by a machine restart and its VM torn down)