Repository navigation
skills/README.md truth + routing sweep: OLD/NEW "evaluate to null since M9.5" is anchored nowhere and contradicts the formula skill; "AI tool params" is not a CEL site; the cross-skill routing list lacks the upgrade-vs-platform and objectui inbound decision points #14794
Description
Activity
Claim: skills lane seat — filed and dispatched in one round (
priority:p3+pm:dispatched;domain:*left to triage, its single producer). The seat dispatches its own programme's follow-up in-lane; #14566 (the same list) is claimed in the same stroke soskills/README.md's routing list is edited once.
Session:session_01LraLgQVGq8egUwfYZpbYt1
Branch:claude/issue-14794-readme-truth-routing
Worktree:objectstack-issue-14794
Clause-②: no — catalog routing prose; item 1 is settled at source by the dev and cited, not claimed. Governedskills/**⇒ draft PR; review requests are the seat's step.Binding on the flight: the two corrections shrink or stay neutral; the three routing bullets (this card's two + #14566's one) are ~25 tokens each in the existing bullets' shape — the #14566 triage's size ruling (comment 5514978050) applies to all three; ⛔ no edit to any
SKILL.md; the generated region of the README untouched; hunks disjoint from PR #14786's:55. First lineFixes #14794andFixes #14566;skip-changeset.
Generated by Claude Code
Claim: dev seat (os-dev subagent of the dispatching PM session — same session id, distinct seat)
Session:session_01LraLgQVGq8egUwfYZpbYt1
Branch:claude/issue-14794-readme-truth-routing(pushed empty at the branch point as a write-route probe)
Worktree:/home/user/objectstack-issue-14794Branch point re-measured:
origin/main89a156af88— identical to the card's measurement (2026-09-03T01:43Z), no drift.Scope taken, exactly as claimed by the lane seat:
skills/README.mdonly — the:87OLD/NEWclause settled at source, the:105"AI tool params" item corrected againstpackages/spec/src/ai, and three routing bullets (this card's two + #14566's one) at ~25 tokens each under the #14566 triage ruling (comment 5514978050). NoSKILL.mdedit, no touch to the README's generated region between the BEGIN/END GENERATED markers, hunks disjoint from PR #14786's anatomy-tree line. One draft PR,skip-changeset, stays draft for human merge.
Generated by Claude Code
os-dev-report
{ "issue": 14794, "also_fixes": 14566, "status": "done", "branch": "claude/issue-14794-readme-truth-routing", "head_sha": "39da552e61a526f10a5951e3cf4d6d07ac78710a", "pr": "https://github.com/objectstack-ai/objectstack/pull/14795", "branch_point": "89a156af88 — re-measured, identical to the card's measurement, no drift. main has since advanced to 8f9ef6f7c8 (4 commits), none touching skills/README.md; PR #14786 has not landed, so the hunks are disjoint and no merge was needed.", "items": [ { "id": "1 — skills/README.md:87", "before": "`@objectstack/spec`. Legacy `OLD` / `NEW` evaluate to `null` since M9.5.", "after": "`@objectstack/spec`. Legacy `OLD` / `NEW` are not CEL — use `previous.` / `record.`.", "source": "False in both halves. (a) M9.5: `git grep -rn 'M9\\.5' -- packages skills docs` returns exactly ONE hit — this README line itself. (b) They do not evaluate to null; they are UNDECLARED IDENTIFIERS. SCOPE_ROOTS (packages/formula/src/cel-engine.ts:95-119) is the published closed set of record-scope roots and contains neither. firstUndeclaredReference (cel-engine.ts:186-205) reports them as bare undeclared references — the class @objectstack/lint raises as an ERROR (packages/lint/src/validate-expressions.ts:490, 'Why it is an error and not a warning'). Measured against the built package, not inferred: SCOPE_ROOTS includes OLD=false NEW=false previous=true record=true; firstUndeclaredReference('OLD.amount != NEW.amount')='OLD'; ('NEW.status')='NEW'; ('previous.amount != record.amount')=null. The replacement's prescription is objectstack-formula's own table, skills/objectstack-formula/SKILL.md:345-346 (OLD.x to previous.x, NEW.x to record.x), so the two published files now agree." }, { "id": "2 — skills/README.md:105", "before": "skill (data validations, automation guards, UI visibility, AI tool params).", "after": "skill (data validations, automation guards, UI visibility).", "source": "Diagnosis confirmed, prescribed replacement falsified. packages/spec/src/ai/tool.zod.ts matches cel|formula|expression ZERO times. But there is no model-registry predicate to swap in: 'predicate' occurs 0 times in packages/spec/src/ai/model-registry.zod.ts; its only expression-typed keys are system/user at :121-122, which are TemplateExpressionInputSchema — defined at packages/spec/src/shared/expression.zod.ts:117-124 as dialect 'template', the {{var}} interpolation form, NOT CEL. skills/objectstack-ai/SKILL.md:405-407 says the same in substance. The AI domain therefore has no CEL site at all, so the dispatch's stated alternative was taken: the item is DROPPED, not replaced. Adding an AI entry to a bullet headed 'Any CEL expression' would have swapped one false claim for another." }, { "id": "3 — skills/README.md:94-110, three routing bullets", "before": "four decision points; none about labels, upgrade-vs-platform, or the consuming UI", "after": "- **Labels vs. bundles** — a `label` you want translated is **objectstack-i18n**. / - **Upgrade vs. platform** — a protocol-major move is **objectstack-upgrade**. / - **Rendering this metadata** — the consuming UI is [objectui `skills/objectui/`](https://github.com/objectstack-ai/objectui).", "source": "Each checked against the TARGET SKILL'S OWN frontmatter description. i18n: 'Author ObjectStack translation bundles — object/field labels, view text …'. upgrade: 'Use when a project is on an older protocol major and must move to the current one, when @objectstack/spec was bumped across a major …'. objectui: path confirmed on the objectui checkout, `git ls-tree origin/main skills/objectui/` at objectui origin/main 39af82f returns SKILL.md, README.md, evals/, guides/, rules/." } ], "premise_false": [ "Item 2's REMEDIATION half is false. The dispatch and the card both instruct: 'The AI domain's real CEL site per skills/objectstack-ai/SKILL.md:405 is a model-registry predicate. Replace the false item with the true one (verify it at packages/spec/src/ai/model-registry.zod.ts — find the predicate field and cite it)'. There is no predicate field: 'predicate' occurs 0 times in that file, and the two expression-typed keys it does carry (system/user, :121-122) are the TEMPLATE dialect, not CEL. The dispatch's own alternative ('or drop the item') was taken. The card's DIAGNOSIS half ('AI tool params is not a CEL site') is confirmed.", "Item 1's original README claim is false in a sharper way than the card states. The card asks whether OLD/NEW 'evaluate to null'; the measured answer is that they are REJECTED as undeclared references — an error raised by @objectstack/lint — not silently nulled. The clause as written described a silent runtime degradation that does not exist." ], "gates": [ { "cmd": "pnpm --filter @objectstack/spec check:skill-docs", "exit": 0, "verdict": "Skill docs in sync (skills/README.md OK) — proves the generated region is byte-identical" }, { "cmd": "node scripts/check-skills-token-ratchet.mjs", "exit": 0, "verdict": "36 authored bundle file(s) within their ceilings; 11 generator-owned file(s) measured, not ratcheted." }, { "cmd": "node scripts/check-skills-token-ratchet.mjs --self-test", "exit": 0, "verdict": "check-skills-token-ratchet self-test: 64 cases pass." }, { "cmd": "pnpm check:role-word", "exit": 0, "verdict": "check-role-word: OK, no new occurrences of the reserved word. 224 files; ledger unchanged at 43 baselined files / 120 occurrences." }, { "cmd": "pnpm check:skill-identifier-liveness", "exit": 0, "verdict": "OK — Leg 1: 465 citation(s) over 46 published file(s); Leg 2: 8 registered exhaustive section(s), 0 ledgered gap(s)." }, { "cmd": "pnpm check:skill-compatibility", "exit": 0, "verdict": "11 SKILL.md file(s) reconciled against 79 workspace packages" }, { "cmd": "pnpm check:published-readme-links", "exit": 0, "verdict": "176 outbound link(s) across 60 published markdown file(s): 0 root-relative, 0 non-canonical origin(s), 27 docs-site page(s) resolved, 103/103 relative target(s) found — the new objectui link resolves" }, { "cmd": "node scripts/check-nul-bytes.mjs", "exit": 0, "verdict": "OK (scanned 8064 text file(s); no raw ASCII control bytes)" }, { "cmd": "node scripts/check-ci-filter-parity.mjs", "exit": 0, "verdict": "union member, green" }, { "cmd": "node scripts/check-cross-package-test-inputs.mjs", "exit": 0, "verdict": "union member, green" }, { "cmd": "node scripts/check-shard-attestation.mjs", "exit": 0, "verdict": "union member, green" }, { "cmd": "node scripts/check-test-completeness.mjs", "exit": 3, "verdict": "PREREQUISITE NOT MET — NOT MEASURED, not red. 'grades a saved turbo run test log, and no log was named … the local reading for this gate is NOT MEASURED.' CI tees the log and passes the path, so CI is unaffected." }, { "cmd": "pnpm --filter @objectstack/lint run check:doc-formula-expressions", "exit": 0, "verdict": "First pass exited 3 (PREREQUISITE NOT MET — imports compiled @objectstack/formula and @objectstack/lint). Both were BUILT and the gate re-run rather than reported as a non-reading, because it is the gate nearest this change (item 1 puts previous./record. into published prose). Then: '22 record-scoped formula example(s) across 426 files / 1365 TS blocks judged clean by @objectstack/formula.'" }, { "cmd": "pnpm --filter @objectstack/spec run check:skill-docs", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:agent-test-spelling", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:corpus-claim-drift", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:cross-package-test-inputs", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:doc-authoring", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:merge-driver", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:pm-governed-merges", "exit": 0, "verdict": "union member, green" }, { "cmd": "pnpm check:skill-frame-sync", "exit": 0, "verdict": "union member, green" } ], "gate_method": "Every exit code captured by redirect BEFORE any pipe (cmd redirected to a log; EXIT read; then tail). Each verdict quotes the gate's own printed line, never a bare exit status. The union was re-derived AFTER the last edit with `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands`; its provenance line reads 'gate list derived from the tree of objectstack-ai/objectstack at commit 39da552e61' with change set '1 path(s) vs merge base 89a156af8' = skills/README.md. All 16 derived commands were run; the 8 dispatch-named gates were run as well.", "reverse_verification": "check:skill-docs proven able to FAIL and to name this file. Implementation committed FIRST, so the restore leg's HEAD already contains it. Guarded by `trap restore EXIT INT TERM`, absolute paths throughout. Mutation landed on disk proven by grep counts flipping BOTH ways inside the generated region ('Construct ObjectQL queries' 1 to 0, 'Construct ObjectQL queried' 0 to 1) and the worktree blob moving 72afcf31b7 to e60bea9ecc — never by an editor's exit code. Mutated leg: exit 1, verdict 'skills/README.md is out of date — run pnpm --filter @objectstack/spec gen:skill-docs'. Restore leg: `git checkout HEAD -- ABSOLUTE_PATH` (pinned to HEAD, never a bare checkout, so it cannot restore from a polluted index), proven by blob hash equality against the HEAD blob 72afcf31b7 (non-empty AND matching), an EMPTY `git diff HEAD`, and anchor counts back to 1 / 0.", "eslint": "Not a narrowing — skills/README.md is outside the linted population entirely, with all three pieces. (1) Population read from eslint's OWN config resolution, not from my reading of the globs: `eslint --print-config skills/README.md` emits an empty document (10 bytes); every files: selector in eslint.config.mjs is {ts,tsx,mts,cts,js,jsx,mjs,cjs} and none matches markdown. (2) Count from --format json for the file: 1 result entry, errorCount 0, warningCount 1, and the single message is eslint's own 'File ignored because no matching configuration was supplied.' — zero rule findings because zero rules apply. (3) Invariance: the repo runs one eslint.config.mjs which 'never enables type-aware linting (no parserOptions.project, no typed @typescript-eslint rules) for ANY file, test or not' (eslint.config.mjs:326-328, carrying its own positive-control measurement), so no untouched file's verdict can move as a function of this diff. A full-repo `pnpm lint` is provably vacuous here.", "token_delta": { "method": "ceil(utf8_bytes / 4)", "file_before": { "bytes": 7941, "tokens": 1986, "at": "89a156af88" }, "file_after": { "bytes": 8237, "tokens": 2060, "at": "39da552e61" }, "file_delta": { "bytes": 296, "tokens": 74 }, "item1": { "before_bytes": 75, "after_bytes": 91, "delta_bytes": 16, "delta_tokens": 4 }, "item2": { "before_bytes": 78, "after_bytes": 62, "delta_bytes": -16, "delta_tokens": -4 }, "items_1_and_2_combined": { "delta_bytes": 0, "delta_tokens": 0, "note": "neutral, as the budget requires — items 1 and 2 fund each other exactly" }, "item3": { "bullet_tokens": [21, 21, 33], "total_tokens": 75, "total_bytes": 296, "ruling": "at the #14566 triage ruling (comment 5514978050): ~25 tokens per bullet, ~75 total" }, "payment": "Item 2's drop of ', AI tool params' (-16 B) exactly funds item 1's truthful replacement (+16 B). A second payment is inside item 3 and is what brought the block from 84 tokens down to the ruling's 75: bullet 2's body does NOT restate 'not objectstack-platform' — the surviving site is the bullet's own heading, 'Upgrade vs. platform', which already names both sides. That is a strict restatement removed, not meaning shaved. NO FURTHER PAYMENT WAS AVAILABLE: the one candidate examined was the process-skill paragraph at :64-66 against the frontmatter rows at :75 (compatibility, 'for a process skill that binds to no schema') and :76 (metadata.domain, 'a delivery-process skill that teaches no schema'); rejected because the paragraph also carries two facts surviving nowhere else — that such a skill 'carries SKILL.md alone' (no references/ directory) and the SKILL_MAP mechanism behind it. Deleting it would have shaved meaning.", "ratchet_population": "skills/README.md is outside it (scripts/check-skills-token-ratchet.mjs:133); the ratchet is green regardless." }, "edit_landed_proof": { "M9.5": "1 to 0", "AI tool params": "1 to 0", "objectstack-i18n": "1 to 2", "objectstack-upgrade": "1 to 2", "objectui": "1 to 2" }, "scope_untouched": "No SKILL.md edited. The README's generated region between the BEGIN/END GENERATED markers untouched (check:skill-docs green proves it byte-for-byte). The anatomy tree at :50-62 untouched, so the hunks are disjoint from PR #14786's :55 — this diff's two hunks start at :85 and :104. No other README sentence edited. Diff is 1 file, +7 -2.", "pr_state": "Draft (draft=true, read back). Label skip-changeset applied and read back as labels=['skip-changeset']. Never flipped ready, no auto-merge, no reviewers requested. skip-changeset justified against scripts/check-empty-changeset.mjs:359-362, which enumerates 'It releases nothing (.github/, .claude/, skills/, docs/, content/, examples/, tests-only, and the like) -to- delete the changeset and apply the skip-changeset label (route 2)', and pins this exact shape as GREEN 3 at :559-565.", "channel": "REST is 403 for this session (curl to api.github.com/repos/objectstack-ai/objectstack returned 403; gh is not installed), so ALL GitHub writes went through MCP — declaring the switch. Label write used the read-union-write-then-read-back fallback; note that issue_read get_labels cannot resolve a PR number ('Could not resolve to an Issue with the number of 14795'), so the read-back was taken via list_pull_requests filtered to this head branch.", "mcp_calls": 15, "out_of_scope_findings": [ "skills/objectstack-ai/SKILL.md:405-407 makes the SAME class of error this PR just fixed in the README, one level down: it writes 'any CEL predicate parses and resolves — in this domain that is a model-registry promptTemplate.system / .user'. Those two keys are TemplateExpressionInputSchema, i.e. dialect 'template' ({{var}} interpolation), NOT CEL — see packages/spec/src/shared/expression.zod.ts:117-124. So a published skill labels a template-dialect expression a 'CEL predicate'. The same sentence also omits the AI domain's other real expression site, the 5-field cron at packages/spec/src/ai/knowledge-source.zod.ts:28. Out of scope here twice over (the dispatch forbids any SKILL.md edit, and #14566's triage forbids expanding sibling skills), and NOT filed per the dispatch's instruction that out-of-scope findings go in the report only. Worth a card at triage's discretion: it is a published-surface truth defect, not cosmetics.", "skills/README.md:114 'Related repositories' still describes objectui only as 'Studio UI (separate repo)' and does not say it publishes a skills catalog. OUI-A-01 noted this; the new routing bullet now carries the route, so the gap is covered functionally, but the two lines are slightly redundant and a future editor may want to merge them. Left untouched deliberately — the dispatch scoped this flight to the routing list and forbade other README sentences." ] }
Generated by Claude Code
ACCEPT — skills-lane seat (session
session_01LraLgQVGq8egUwfYZpbYt1), reviewer of record, verified against GitHub and the seat's own checkout, not the report. Covers #14794 and #14566 (one PR, bothFixes).- PR docs(skills): settle two false claims in skills/README.md and add three cross-skill routing bullets #14795 exists, is DRAFT, head
39da552e, one file, +7 −2; labelskip-changeset; noneeds:contract-review(catalog routing prose; item 1 cited at source, not claimed). - Diff read by the seat, item by item: (1)
:87Legacy OLD / NEW evaluate to null since M9.5.→Legacy OLD / NEW are not CEL — use previous. / record.— seat re-readpackages/formula/src/cel-engine.ts:95-119at8f9ef6f7c8:SCOPE_ROOTSis the closed root set (record,previous,input,output,os,vars, …,current_user) and carries neitherOLDnorNEW;firstUndeclaredReference(:77,:186-205) reports them as undeclared — the lint's error class, not a silentnull.M9.5was anchored nowhere inpackages/orskills/. The new clause matchesskills/objectstack-formula/SKILL.md:345-346. (2):105drops, AI tool params:tool.zod.tshas no expression field, and the seat's own brief was wrong that a "model-registry predicate" is the AI domain's CEL site —model-registry.zod.ts:121-122isTemplateExpressionInputSchema, dialecttemplate(expression.zod.ts:117-124); the dev took the brief's drop branch correctly, and the same false claim inskills/objectstack-ai/SKILL.md:405-406is now [finding]skills/objectstack-ai/SKILL.md:405-406calls a model-registrypromptTemplate.system/.user"a CEL predicate" — those keys are thetemplatedialect ({{var}}), and the AI domain has no CEL site at all #14797. (3) three routing bullets at:110-113— labels vs bundles → objectstack-i18n ([finding] Nothing routes an agent authoring object/field/view labels toobjectstack-i18n— the cross-skill routing list inskills/README.mdhas no labels-vs-bundles decision point #14566, verified against that skill's frontmatter), upgrade vs platform → objectstack-upgrade, rendering this metadata → objectuiskills/objectui/(path verified in the objectui checkout). - Seat re-measured at 2026-09-03T02:11Z:
M9.51 → 0,AI tool params1 → 0,objectstack-i18n/objectstack-upgrade/objectui1 → 2 each; hunks at:87,:106,:110-113— disjoint from PR docs(skills): admit hand-authored reference files underreferences/in the catalog anatomy #14786's:55; the generated region between the markers byte-identical (check:skill-docsgreen; ablation red-then-restored with blob hashes). - Budget: 7,941 → 8,237 bytes = 1,986 → 2,060 tokens (+74). Items 1 + 2 net 0 (they fund each other exactly, +16 / −16 B); item 3 is 21 + 21 + 33 = 75 tokens — at the [finding] Nothing routes an agent authoring object/field/view labels to
objectstack-i18n— the cross-skill routing list inskills/README.mdhas no labels-vs-bundles decision point #14566 triage's size ruling (comment 5514978050), the three bullets in the existing four's shape. The file is outside the ratchet population; the one further payment candidate (:64-66) was examined and rejected because it carries two facts surviving nowhere else — the seat agrees. - CI at
39da552e: every one of the 38 check runs is success or skipped, read at 2026-09-03T02:32Z —Lint & Repo Gatessuccess, the fourType Checklegs, all sixTest Coreshards,Check Changesetsuccess,Governed Surface Queue Guard(draft — correct),Check Documentation Links, the three claim guards. - Recorded on the ledger, not filed:
skills/README.md:114"Related repositories" now slightly overlaps the objectui bullet — left alone by dispatch. - Governed
skills/**⇒ stays DRAFT; review requested from os-zhuang + hotlong; human merge is the record. At MERGED the seat stripspm:dispatchedon both skills/README.md truth + routing sweep:OLD/NEW"evaluate tonullsince M9.5" is anchored nowhere and contradicts the formula skill; "AI tool params" is not a CEL site; the cross-skill routing list lacks the upgrade-vs-platform and objectui inbound decision points #14794 and [finding] Nothing routes an agent authoring object/field/view labels toobjectstack-i18n— the cross-skill routing list inskills/README.mdhas no labels-vs-bundles decision point #14566 (Fixes⇒ both close).
Generated by Claude Code
- PR docs(skills): settle two false claims in skills/README.md and add three cross-skill routing bullets #14795 exists, is DRAFT, head
Merge patrol reading 2026-09-03T09:10Z (
git fetch origin main, thengit log origin/main --grep '(#14795)'in the same minute): PR #14795 merged asf22ed22c24at 02:35:08Z, landing this card and #14566 together; this card was closed by os-zhuang at 04:12:17Z. The seat's patrols between 02:35Z and 08:31Z kept listing PR #14795 as a draft at the governed terminal — wrong from 02:35Z on; corrected here and in seat post #7623 at this refresh.pm:dispatchedremoved at 09:12Z; labels nowdocumentation,priority:p3, read back 09:13Z.
Generated by Claude Code
Filed by the skills lane seat (session
session_01LraLgQVGq8egUwfYZpbYt1) from the optimization programme's audit ledger (#14292: UPG-A-01, OUI-A-01, and two README truth items the per-package flights could not take becauseskills/README.mdwas outside every flight's file surface). Bare, unassigned; the seat dispatches it in-lane. Sibling: #14566 (I18N-A-03, the labels-vs-bundles bullet on the same list) — the same flight lands both, one PR, so the routing list is edited once.Measured at
origin/main89a156af88, 2026-09-03T01:43Zskills/README.md:87— "LegacyOLD/NEWevaluate tonullsince M9.5."M9.5occurs nowhere else inpackages/orskills/;git grep -E '\bOLD\b|\bNEW\b'overpackages/formula/src(non-test) returns 0 — no code evaluates those spellings tonull. The formula skill says the opposite in substance:skills/objectstack-formula/SKILL.md:345-346mapsOLD.x→previous.xandNEW.x→record.xas legacy spellings to migrate, and:361statespreviousis bound only on the events listed there. Two published files disagree; the README's sentence has no anchor. Settle at source (what the CEL parser actually does with a bareOLD) and keep one true clause — most likely "legacyOLD/NEWare not CEL; migrate per objectstack-formula's table" or simply drop the clause.skills/README.md:105— the "Any CEL expression" bullet lists "AI tool params" as a CEL site.git grep -i 'cel|formula|expression'overpackages/spec/src/aifinds CEL nowhere on tool params — only{{var}}template interpolation on model-registry prompts (model-registry.zod.ts:121-122). The AI domain's real CEL site, perskills/objectstack-ai/SKILL.md:405, is a model-registry predicate. Replace the false item with the true one (or drop it).:94-110, four bullets) lacks two decision points the audits measured as real load-failure modes, in the same shape [finding] Nothing routes an agent authoring object/field/view labels toobjectstack-i18n— the cross-skill routing list inskills/README.mdhas no labels-vs-bundles decision point #14566 records for i18n:objectstack-platform(kernel/plugin lifecycle) and never learnsobjectstack-upgrade(:42in the index) is the process skill for that; one bullet routing "moving a metadata project across a protocol major / a breaking changeset" to objectstack-upgrade.:114); an agent authoring schema-driven UI for an app never learns theobjectuiskill (separate repo,skills/objectui/) exists; one bullet routing "rendering / composing the UI that consumes this metadata" to the objectui catalog.objectstack-i18n— the cross-skill routing list inskills/README.mdhas no labels-vs-bundles decision point #14566 — the same list, landed by the same PR.Scope and budget
Fixes #<this>andFixes #14566. Items 1–2 are corrections (shrink or neutral). Items 3's three bullets are funded additions of roughly +25 tokens each, in the four existing bullets' shape and size — ⛔ no more than that (maintainer 2026-08-21: 「……不能为了一个小功能扩写很多。」); pay inside the file where a strict restatement exists, net as close to 0 as honesty allows, reported per item.skills/README.mdis outside the token-ratchet population (check-skills-token-ratchet.mjs:133); the generated region between<!-- BEGIN/END GENERATED: skills -->is untouched (check:skill-docs --checkstays green). PR docs(skills): admit hand-authored reference files underreferences/in the catalog anatomy #14786 (at the governed terminal) holds one line of the same file (:55, the anatomy tree) — disjoint hunks; the later to land mergesmainin.skills/**⇒ draft, human merge;skip-changeset.Refs: #14292 · #14566 · #14315 (I18N flight) · #14307 (UPG flight) · objectui#7251 (OUI flight).