Repository navigation
v1.91.284.0 feat(absorb-tstack): /modernize skill, per-skill Codex default prompts, gstack-post --attribute, the unattended grant and a CONTRIBUTING porting section - #3497
Merged
Conversation
…s, gstack-post --attribute, the unattended grant and a CONTRIBUTING porting section Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…re the context ceiling Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
|
😎 This pull request was merged. |
- gstack-post: with --attribute, report scan findings at the body file's own line numbers and a finding inside the callout as PART: attribute (the coordinates were shifted by the three callout lines) - /modernize: route commit messages through gstack-issue-guard with the rest of the PR text, read isCrossRepository and maintainerCanModify, fetch a fork head from pull/N/head, refuse a PR whose head is the base branch, and list moved files NUL-safely - docs/unattended.md: say that /ship never rebases or force-pushes and an unattended /modernize stops at its verdict - CHANGELOG: seven skills carry a codex_default_prompt, not six Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…hen the PR moved A rerun after the author force-pushed left refs/modernize/pr-N stale (the non-forced refspec is rejected with exit 0 through the pipe), and the lease pinned to the current remote head would then have let the rebased stale head overwrite the author's newer commits. The fetch is forced, Steps 2 and 4 stop when the head is not the one observed in Step 1, and the head-is-base guard applies only to same-repository PRs (a fork's main into main is a valid head). Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…ork head to the fork A plain git rebase drops merge commits and git range-diff compares only the non-merge patches, so a change authored in a PR's merge commit disappeared while the proof printed = for every patch and Step 5 force-pushed the loss. Step 4 now counts merges in $MERGE_BASE..$OLD_HEAD and stops when there are any, switches with -C so a rerun after a refused lease does not fail on the existing branch, and asks for a cross-repository diff to be read as untrusted input before the repository's test command runs it. Step 5 pushes a fork head to the fork's URL with the same pinned lease instead of to origin, where the lease could never match. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…p proof and the unattended grant Adds the /ship coverage audit's tests: test/host-renders-openai-yaml-proof.test.ts drives bin/gstack-host-renders.sh's weakened proof against a rendered yaml, a hand-written one and a generic-prompt one; test/unattended-grant.test.ts pins the docs/unattended.md grant section and the handoff pointer; test/gen-skill-docs.test.ts adds modernize to the skills whose codex_default_prompt must reach openai.yaml. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…adata file From the ship documentation audit: CONTRIBUTING's new-skill checklist said no static list needs a row, which this PR disproved (lib/doctor-components.ts SKILL_REQUIREMENTS and the catalog rows); the host comparison table and docs/ADDING_A_HOST.md gain the agents/openai.yaml default_prompt source; a blank line before the new docs/unattended.md grant heading. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
GSTA-36 rule-19 grant 2026-10-27. CHANGELOG via mergeChangelogs; lib/gstack-post.ts and its test keep both --replace (#3483) and --attribute; catalog ratchet re-measured on top of design-motion: 1,314 -> 1,359 (5,427 bytes = 1,357; modernize +178); context-budget.json modernize 9466 and alwaysOnTotal 7880 recaptured; bin/gstack-doctor-components.sh regenerated. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
CHANGELOG [Unreleased] gate-integrity record: the catalog and context-budget figures re-measured on top of main (1,314 -> 1,359; modernize 9466; alwaysOnTotal 7458 -> 7880), as the merge commit's ratchets state. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…cklist Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
GSTA-36 rule-19 grant 2026-10-27 naming 76aa68f. CHANGELOG via mergeChangelogs; docs/unattended.md keeps both appended sections, main's 'The decision trail' first, this PR's 'What an unattended session may do' after it; nothing else hand-resolved. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
GSTA-36 rule-19 grant 2026-10-27 naming f27a760. CHANGELOG via mergeChangelogs; nothing else hand-resolved. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
GSTA-36 grant 2026-10-27. On the tree merged with f27a760 the ship union is 326,691 / 187,706 = 1.7404 against #3487's 1.7400; the only ship bytes this PR adds are the handoff section's one-line pointer. Cap 1.7400 -> 1.7405, measurement appended to the comment. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…p cap re-pin Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
GSTA-36 rule-19 grant 2026-10-27 naming 1f246e3. CHANGELOG via mergeChangelogs ([Unreleased] on top); CONTRIBUTING.md keeps this PR's new-skill steps 3-4 and main's step 5 (generated sections/preamble.md); test/fixtures/context-budget.json keeps main's values with modernize recaptured at 3772 (the preamble hoist moved it from 9466) and alwaysOnTotal 7769 -> 7880; test/helpers/carve-guards.ts ship cap 1.7524 -> 1.7528, measured 329,005 / 187,706 = 1.75277 on the merged tree (skeleton 60,900 after the hoist, 0 bytes from this PR); modernize/sections/preamble.md generated by the hoist. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
CHANGELOG [Unreleased]: end the Codex default-prompts sentence at "generator input only." so bin/gstack-prose-lint --unreleased --gate (main's new quality-gate step) passes, and restate the gate-integrity bullet's figures for the tree merged with 1f246e3: ship cap 1.7524 -> 1.7528 (329,005 / 187,706), modernize 9466 -> 3772 after the preamble hoist, alwaysOnTotal 7769 -> 7880, catalog 1,359 unchanged. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
GSTA-36 rule-19 grant 2026-10-27 naming 392f50d. CHANGELOG via mergeChangelogs (released entries main's, this PR's entry under [Unreleased]); test/helpers/carve-guards.ts ship cap on main's line 1.7525 -> 1.7529, measured 329,024 / 187,706 = 1.75287 on the merged tree (skeleton 60,900, 0 bytes from this PR). Nothing else hand-resolved. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
CHANGELOG [Unreleased] gate-integrity bullet names the current ship cap pin (1.7525 -> 1.7529, measured 329,024 / 187,706 on the 392f50d merge) and keeps the 1.7528 pin in the earlier-merges clause. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
MERGE_CONDITION: sha=61a145c5ec19ebeacca828498247e0f54f2aa363 NEW_VERSION=1.91.284.0 base=215c76e334a071441f98a36e1806f45526751663 pr=3497 order=3497 GSTACK_RESULT: skill=restamp status=complete run=/workspace/gstack/.gstack/tmp/restamp-journal.json 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
…cut-over Rule 19 grant (GStack Minder) for one merge. carve-guards keeps sizeLedger for ship; main's maxSizeRatio 1.7529 re-pin (#3497) moved verbatim to the top of test/size-ledger/ship/HISTORY.md. base.json re-measured on the merged tree: 329,024 union bytes. CHANGELOG released entries are main's byte for byte. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
9 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.
What changed
Before: every skill's
agents/openai.yamlcarrieddefault_prompt: "Use <name> for this task.". After: a template'scodex_default_prompt:key is what Codex's picker inserts; ship, review, qa, autoplan, office-hours, land-and-deploy and modernize carry one.New
/modernizeskill: one of three verdicts is written before anything is pushed; CLOSE is a recommendation the skill never carries out itself.Text version
git range-diff, the repo's tests,--force-with-lease=<branch>:<observed sha>Five items from
plans/absorb-tstack.md§4 (items 3, 5, 6, 7, 8), each written in gstack's words; no sentence is copied from tstack (NOTICE.md entry extended).scripts/gen-skill-docs.ts):extractCodexDefaultPromptreads a template'scodex_default_prompt:key andgenerateOpenAIYamlemits it, falling back to today's string. The key is generator input only: Claude's denylist strips it (hosts/claude.ts), the allowlist hosts never copy it, so every tracked SKILL.md render is byte-identical (git statusaftergen:skill-docs --host allshows no render change except the new skill and the root router's one routing line; the ship skeleton is unchanged at main's value (82,645 bytes then, 82,697 after f27a760) and no entry intest/fixtures/skill-md-size-allowlist.jsonmoved).bin/gstack-host-renders.sh's weak proof that a yaml is gstack's now accepts anydefault_promptline. There are noagents/openai.yamlgoldens undertest/fixtures/golden/andscripts/preflight-codex-overlap.tschecks paths, not bytes, so neither needed a change; the assertions live intest/gen-skill-docs.test.ts./modernize(modernize/SKILL.md.tmpl, tier 2, no sections): reads the PR as a claim (title, body, comments and commit messages only throughgstack-issue-guard), reads what moved under it for conflicts in meaning, writes the verdict, then for MODERNIZE rebases, proves intent withgit range-diffand pushes with a lease pinned to the observed head. The /ship review passes hardened it: a fork PR's head is fetched frompull/N/headwith a forced refspec and every step stops when the head is not the one observed; Step 4 refuses a PR range that holds merge commits (a plain rebase drops their changes andrange-diffcannot show it); Step 5 pushes a fork head to the fork's URL with the same lease./carefulalready rates the lease push MEDIUM (--force-with-leaseis never HIGH), so it is unchanged. Rows in README, AGENTS.md, docs/skills.md (table and deep dive), docs/PROJECT_STRUCTURE.md, the root router,lib/doctor-components.ts;gstack/llms.txtandbin/gstack-doctor-components.shregenerated.Co-authored-by; yours on top; the original closed with a specific thank-you linking the port);docs/captain.mdpoints at it. No/takeoverskill. The ship documentation audit also corrected the "Adding a new skill" checklist (the static lists a new skill still needs) and added the Codex metadata-file row to the host table and todocs/ADDING_A_HOST.md.gstack-post --attribute "<model> on behalf of <login>"onpr-comment,issue-commentandreply(lib/gstack-post.ts): opt-in, off by default, prepends> [!NOTE]/> 🤖 **<text>**. The callout bytes are scanned and bound into the MEDIUM confirmation token; findings in the body are reported at body-relative lines and callout findings asPART: attribute; refused on the other operations and for empty, multi-line or over-200-character text.review/greptile-triage.mdgains two lines saying when to add it (GSTA-113 granted the edit).docs/unattended.md"What an unattended session may do" (new last section): merge never; force-with-lease only on the run's own branch after a named rebase, withgit range-diffand a lease pinned to the observed sha; /ship itself never rebases or force-pushes and an unattended /modernize stops at its verdict; tests only by the methods the repo's instructions name, no ask-first; proof is what the run saw.ship/sections/handoff.md.tmplgains a one-line pointer: 87 bytes in the lazy section, 0 skeleton bytes.Merge. Commit 4cd14f9 merges origin/main (v1.91.267.0) into the branch under GStack Minder's rule-19 grant of 2026-10-27 (the grant is spent): CHANGELOG merged additively by
mergeChangelogs,lib/gstack-post.tsand its test keep both main's--replace(#3483) and this PR's--attribute, the catalog ratchet andtest/fixtures/context-budget.jsonwere re-measured on top of main's/ubicloudand/design-motionrows, andbin/gstack-doctor-components.shwas regenerated. The second /ship run's documentation audit then named the optionalcodex_default_prompt:key in CONTRIBUTING.md's new-skill checklist (0b2a7c8). Commit 58f7259 merges origin/main again (v1.91.268.0, 76aa68f) under GStack Minder's second rule-19 grant (also spent): CHANGELOG merged bymergeChangelogswith[Unreleased]kept on top, anddocs/unattended.mdkeeps both appended sections, main's "The decision trail" first and this PR's "What an unattended session may do" after it; no prose changed, andgen:skill-docsis fresh. Commit 9b1bd02 merges origin/main a third time (v1.91.270.0, f27a760) under GStack Minder's third rule-19 grant (also spent): CHANGELOG merged bymergeChangelogswith[Unreleased]kept on top was the only hand resolution; commit 50e6428 then re-pins the ship union cap intest/helpers/carve-guards.ts(Budgets below), the one change since the third /ship receipt, which the carry check moved to this head. Commit 14fb00c merges origin/main a fourth time (v1.91.278.0, 1f246e3, the shared-preamble hoist) under GStack Minder's fourth rule-19 grant (also spent), with four hand resolutions: CHANGELOG bymergeChangelogswith[Unreleased]on top (reproduced byte for byte by QA), CONTRIBUTING.md keeps this PR's new-skill steps 3-4 and main's step 5,test/fixtures/context-budget.jsonkeeps main's rows withmodernizerecaptured at 3772 andalwaysOnTotal7880, and the ship cap re-pinned at 1.7528 (Budgets below);modernize/sections/preamble.mdis the hoist's generated output. Commit ae4b9fd is the fifth /ship run's review fix: two CHANGELOG sentences (the prose-lint colon the Quality gate failed on, and the gate-integrity figures brought to the merged tree). Commit 5b395f1 merges origin/main a fifth time (v1.91.281.0, 392f50d, which landed #3515 and #3516's review-army edits) under GStack Minder's fifth rule-19 grant (also spent), with two hand resolutions: CHANGELOG bymergeChangelogswith[Unreleased]on top and the released entries main's byte for byte (reproduced by QA), and the ship cap re-pinned at 1.7529 (Budgets below). Commit 61a145c is the sixth /ship run's review fix: the CHANGELOG gate-integrity bullet's cap figures brought to this merged tree.Why
Garry asked what tstack does and what gstack should take. These five are the takes that needed no new mechanism. A maintainer with aging contributor PRs had no gstack verb for "is this still worth landing"; Codex users saw "Use Ship for this task." in the picker; an unattended /ship had no written statement of what it may do with git and tests; the porting rule lived only in a Captain's drive; and gstack-posted comments carried no author line when a maintainer wanted one.
Approvals
Source: gstack-approvals record for absorb-tstack-2 (display only; not an approval citation).
Summary
/modernizeskill and its catalog rows, hardened by the review passes (fork-head refspec and moved-head stops, merge-commit guard, fork push target).codex_default_promptrendered intoagents/openai.yaml; every SKILL.md render byte-identical.gstack-post --attributeopt-in author callout with body-relative scan coordinates.docs/unattended.md, the CONTRIBUTING porting section, and the documentation-audit corrections to CONTRIBUTING.md anddocs/ADDING_A_HOST.md.test/modernize-skill.test.ts,test/host-renders-openai-yaml-proof.test.ts,test/unattended-grant.test.ts(new);test/gen-skill-docs.test.tsandtest/gstack-post.test.tsextended.Budgets and gate integrity
test/helpers/carve-guards.ts: the ship union capmaxSizeRatio1.7525 → 1.7529 (commit 5b395f1, a budget change made under GStack Minder's fifth written rule-19 grant; earlier in this PR the same line pinned 1.7405 on top of v1.91.270.0 feat(ship): version-less repos, docs-audit record and caps, the frozen-scope paid case, and the review-loop follow-ups #3487's 1.7400 and 1.7528 on top of main's 1.7524). Main pinned 1.7525 at a measured 1.7524 after v1.91.281.0 fix(flaky): repair four paid gate detectors, the design-floor focus picker and the judged review/plan-review prose (GSTA-128) #3516 (the GSTA-128 flaky pass); on the tree merged with 392f50d (v1.91.281.0) the union is 329,024 / 187,706 = 1.75287, measured withreadSkillForParity('ship')(QA capture 003), and this PR's only ship bytes are the handoff section's one-line pointer (87 lazy bytes). The ship skeleton is main's 60,900 bytes with 0 bytes from this PR.bun test test/parity-suite.test.tson this tree: 2 pass, 0 fail.test/fixtures/context-budget.json:modernize9466 → 3772 (the hoist moved its preamble into a generatedsections/preamble.md, so the always-loaded part shrank) andalwaysOnTotal7769 → 7880 (main's value plus the /modernize router line), both recaptured on the merged tree; every other row keeps main's committed value, includinggstack5350 where a fresh capture would write 5384 (the committed value is the stricter one; the ratchet passes at 4,895 ≤ 5,350).test/catalog-budget.test.ts: 1,359 token-equivalents, unchanged by the merge; measured 5,427 bytes = 1,357 on the merged tree (2 left). CLAUDE.md, CONTRIBUTING.md and docs/CONTRIBUTOR_REFERENCE.md repeat the number.test/fixtures/skill-md-size-allowlist.json, timeout, trial count or case kind changed; the cap re-pin above is the one threshold this PR moves.Test Coverage
Coverage audit: the sixth /ship run's material (a rule-23 delta pass at 5b395f1, then 61a145c) is a test-helper threshold line and CHANGELOG text, so the audit returned skipped (no new application code paths to audit; coverage null, 0 gaps). The fifth run's audit-only result stands for the unchanged product paths (the carry receipt records the same product patch-id): 94% value-weighted; 15 of 16 paths rated ★★ or better, 0 weakly covered, 0 gaps, gate PASS (target 80%).
Tests: 1504 → 1504 (+0 new) (the first run added 2: 1481 → 1483; main has grown since).
Coverage: 94% value-weighted (100% including 0 weakly covered paths), reused from the fifth run
Test value: 0 tests written, 2 rejected by the authoring gate, 0 existing tests extended, 0 paths weakly covered (weak = ★, gate-failing or unrated).
Remaining gap:
[->EVAL] /modernize behavior—modernize-skill.test.tspins the prose bytes; whether the skill writes a verdict before any push and never closes a PR is agent behaviour, covered only by a paid behavioural eval (see Eval Results). Proposed eval card:Value: protects=/modernize writes a verdict before any push and never closes a PR; fails_when=skill pushes or closes before disposition; why_new=modernize-skill.test.ts pins prose only; seam=none; no_claim=model behaviour on real PRs beyond the fixture.Tests ran: typecheck count unavailable
Tests ran: parity 2
Tests ran: focused 563
Tests ran: remote gate (Ubicloud,
gstack-pregate --stage tests --remote, run ship-7, tree a3f8f004) 21056Pre-Landing Review
Sixth /ship run, after the fifth rule-19 main merge. The carry check carried the fifth run's receipt to 5b395f1 (base merge, a test-only change and release metadata), and the Captain asked for a fresh receipt at this head, so Steps 2-16 ran as a rule-23 delta pass. Pass 1 (
git diff ae4b9fdd9..5b395f15c -- test/helpers/carve-guards.ts CHANGELOG.md, 59 lines, the full diff vs main as context): the two hand resolutions verified against origin/main (carve-guards differs from main in the shipmaxSizeRatioline only, re-measured 329,024 / 187,706 = 1.75287 with skeleton 60,900; CHANGELOG's[Unreleased]block is byte-identical to ae4b9fd and its released entries byte-identical to origin/main, withmergeChangelogsreproducing the committed file). Core checklist: 1 finding, fixed in 61a145c — CHANGELOG.md:20 the gate-integrity bullet still quoting the previous pin (INFORMATIONAL; rewritten to 1.7525 → 1.7529, 329,024 / 187,706 on 392f50d, main's 1.7525 at a measured 1.7524 by #3516, with the 1.7528 pin moved into the earlier-merges clause). Gate Integrity: the cap hunk is justified by the measurement visible in the diff and reproduced here, under the Captain's grant; no gate finding. Specialists testing, maintainability, performance and state-machines (STATE_MATRIX_CHECK: ok machines=0): all NO FINDINGS (security, simplification and the red team not activated: 59 lines). Pass 2 (range 5b395f1..61a145c, 2 lines):PROSE_LINT: ok, figures re-verified against the tree, state-machines NO FINDINGS, clean and converged. Approval classes: none. The earlier runs' passes are in this PR's earlier body revisions.REVIEW_LOOP: passes=2 blocking_found=0 deferred=0 noted=0 fixed=1 scope=frozen
Gate edits: none detected (12 listed, 5 read; patterns: RH-1,RH-3,RH-4,RH-12,RH-13,RH-14,RH-15,RH-16; unlisted idioms: none)
Outside review: Codex CLI is not installed on this machine (
CODEX_MODE not_installed), so the Codex adversarial and structured (1,248-line diff) reviews are missing coverage. In their place, on this run's range and delta: a Capy outside-voice subagent (openai/gpt-6.1-sol, platform record agrees) returned NO FINDINGS and recommended proceeding ("the scoped changelog preserves released entries and the documented, authorized cap re-pin covers the supplied measurement"), and the native adversarial pass (anthropic/claude-fable-5-1) returned NO FINDINGS after reproducing the measurement withreadSkillForParityand the parity suite and checking the released CHANGELOG entries against origin/main byte for byte ("ship because the range is one CHANGELOG bullet plus a test-helper cap re-pin whose measurement I reproduced").Review notes
Below the blocking floor (advisory, informational, test-gap or wording items); recorded here and not fixed in this PR. Both were raised by the fifth run's Step 11 native adversarial pass and still describe the tree; this run's adversarial passes returned NO FINDINGS, so
REVIEW_LOOPabove countsnoted=0.CHANGELOG.md:20The gate-integrity bullet saystest/fixtures/context-budget.json'is recaptured on the merged tree'; a recapture at HEAD would also writegstack5384 (the root router grew by the /modernize line) while the fixture keeps main's 5350, so only the two rows are taken from a recapture and thegstackrow deliberately keeps main's stricter value. The ratchet passes either way (4,895 ≤ 5,350).test/helpers/carve-guards.ts:171The shipmaxSizeRatiocomment's lead clause still justifies main's pin (now v1.91.281.0 fix(flaky): repair four paid gate detectors, the design-floor focus picker and the judged review/plan-review prose (GSTA-128) #3516's 1.7525, 'measured 328,937 / 187,706 = 1.7524'); this PR's 1.7529 measurement (329,024 / 187,706) sits as the last clause of the line, unlike the file's convention of current measurement first and history after 'Before:'. Comment only; no test reads it.Exploratory QA
Functional smoke on the merge (report-only,
.gstack/qa-reports/ship-3497-20261027T195603Z/functional/, owned fixture with CHANGELOG snapshots fromgit show; no network, no GitHub call): 5 captures, verdict pass —gstack-prose-lint CHANGELOG.md --unreleased --gateok;mergeChangelogs(ours@ae4b9fdd9, theirs@392f50d4, base@1f246e35)equals the committed CHANGELOG.md byte for byte (2,096,435 bytes) withcheckChangelogclean;readSkillForParity('ship')329,024 / 187,706 = 1.75287 under the 1.7529 cap, skeleton 60,900;bun test test/parity-suite.test.ts2 pass;bun test test/changelog-merge-damage.test.ts test/skill-md-size-guard.test.ts8 passing tests. The pass-2 smoke (ship-3497-20261027T195945Z, 4 captures, verdict pass) re-ran the prose-lint gate on the rewritten bullet, proved the edit additive (released entries equal the pure merge; the only line differing from it is line 20, which carries the current pin), re-checked the figures against the tree and graded the parity and changelog-damage ratchets on the committed tree. Not exercised: the live posting path, GitLab, the paid /modernize behaviour eval. The earlier runs' smokes stand for the earlier trees.Design Review
No frontend files changed — design review skipped.
Eval Results
No paid eval ran.
bun run eval:select(pr profile) reports a full fallback — 117 behavioral probes + 24 quality judges restored becauseNOTICE.md,scripts/gen-skill-docs.tsandscripts/host-config.tschanged — and this unattended run was told not to start one; it is a pending approval item for GStack Minder (the /modernize behaviour has no paid case yet). CI E2E Evals (pr-summary slice) runs on every push of this PR; its result at 61a145c is in the checks.Scope Drift
Scope Check: CLEAN (every change is one of the five plan items, its tests, one of the five rule-19 main merges, or a review/documentation fix inside them).
Plan Completion
Plan completion audit: not run (no plan is bound to this branch and no docs/designs/ file matches). Fix: add "Plan: " to the PR body, or run /autoplan.
Verification Results
No plan-specific verification obligations (VERIFY_RESULT=skipped).
TODOS
No TODO items completed in this PR.
Documentation
Status: current. No edits needed; every checked claim matches the tree at head 61a145c (base 392f50d).
Scope: 43 changed files from the candidate (committed only; none staged, unstaged or untracked) plus the authored docs they reference, checked against catalog-budget, modernize template, gen-skill-docs/host configs, handoff template, captain.md and ship-policy.md.
Documentation health:
## /modernizeanchor)Coverage debt: none found for /modernize (all five catalog surfaces plus lib/doctor-components.ts row present). Diagram drift: none; no diagrams touched.
Test plan
bun run typecheck: exit 0 (tree a3f8f004, 19:58Z)bun test test/parity-suite.test.ts: 2 passbun test test/gstack-post.test.ts test/modernize-skill.test.ts test/catalog-budget.test.ts test/gen-skill-docs.test.ts test/unattended-grant.test.ts test/host-renders-openai-yaml-proof.test.ts test/context-budget-ratchet.test.ts test/skill-md-size-guard.test.ts test/doctor-components.test.ts test/instruction-facts.test.ts test/changelog-merge-damage.test.ts: 563 pass, 0 fail (11 files)bun run gen:skill-docs --host all: exit 0, no render change (fresh)bun --no-env-file --config=/dev/null bin/gstack-prose-lint CHANGELOG.md --unreleased --gate:PROSE_LINT: ok files=1gstack-pregate --stage tests --remote --spend-cap 2 --timeout 2700 --allow-repo-commands --run ship-7on tree a3f8f004 (Ubicloud): 21,056 tests, 0 failed, $0.07 (the parent's run on 5b395f1, tree 2c83ad26: 21,056 tests, 0 failed, $0.07)gstack-pregate verifyok🤖 Generated with Claude Code
Greptile Review
No Greptile comments.
Issues
Refs #3483
Refs #3487
Blast radius
The generator change (
scripts/gen-skill-docs.ts,hosts/claude.ts) feeds every host render andagents/openai.yaml;bun run gen:skill-docs --host allon this tree leavesgit statusclean, so no tracked SKILL.md render moved (script ran at Step 16 of this run, tree a3f8f004).lib/gstack-post.tsis reached by every gstack-posted comment; without--attributethe body is sent exactly as written (test/gstack-post.test.ts, 563-test focused lane green at this tree), and the flag is refused onpr-body, so existing callers see no change.