docs(skills): retirement kit routes a RETIRED_KEYS_BY_MAJOR entry through entries/retired-keys/ and gen:migration-registry, and an enum-value narrowing through enumWithRetiredValues()

[objectstack-fleet]

Fixes #20803

Clause-②: no

Dispatched by the domain:skills seat 1 (seat post #7623) under the PM claim 5907960440, session session_01KTZmMfzVzjNvyaLyQ8mHvg, branch claude/issue-20803-retirement-kit-registry-entries. This is a draft. The file is on a governed surface (.claude/**, Tier S in the GOVERNED_SURFACES register), so the seat lands it after an at-tier contract review.

维护者速读(草稿)

• 改了什么:退役技能包 .claude/skills/spec-property-retirement/SKILL.md 的两段文字。§3 把「退役键登记」改成在 entries/retired-keys/ 加一个文件、再跑 gen:migration-registry;§2 的枚举值退役改指现成的 enumWithRetiredValues(),不再教手写 error map。文件行数不变(337/337)。
• 为什么改:登记表 registry.ts 早已改为由 entries/ 逐文件生成,旧文字却让照做的 agent 往生成区手打,check:migration-registry 必红、白费一轮;枚举值退役已有通用 helper,旧文字仍在教旧的手写形状。
• 风险与代价(含回滚):纯文字改动,零运行时影响,不发布任何包。回滚即 revert 本 PR。它与 spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809(门 (b) 打印的补救文案)须说同一条路线:本 PR 先落,spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 引用本文。
• 席位意见:
• 你要做的:无。本 PR 属 Tier S,由席位达档复核 PASS 后入队落地。

What changed

One file, .claude/skills/spec-property-retirement/SKILL.md, two texts, net 0 lines (337 before and after; ceiling 337). The landing site is the one the claim named. The measurement found no other producer of the wrong text inside .claude/**.

Text 1: §3, the RETIRED_KEYS_BY_MAJOR checklist item. It sent the entry to be typed into packages/spec/src/migrations/registry.ts, and it said the gate's failure output prints "the line to paste". That table is now generated: it is concatenated from one file per entry, and the generator reverts a hand edit between the markers, which check:migration-registry reports first. The item now says what an author actually does:

• add one file under packages/spec/src/migrations/entries/retired-keys/, named as entries/README.md says. The kit does not restate the naming convention.
• run pnpm --filter @objectstack/spec gen:migration-registry, which generates the entry into registry.ts. Never type the entry there.

Kept unchanged:

• the literal 'DEFKEY:NAME' value, spelled as authorable-surface/CATEGORY.json spells it (minus [RETIRED]), filed under the current major;
• gate (b) reads it by exact set membership, with no inference;
• the (b2) warning: do not add the entry before the tombstone lands;
• the "why a second table" note and the aging-clock note.

Two lines of the old item are byte-identical in the new one: the authorable-surface/… line and the (b2) tail line.

Deleted: "the gate's failure output prints the line to paste". That sentence named the paste route this card retires, and what gate (b) prints is #20809's to change. The new text is true both before and after #20809 lands, because it names the file route and never quotes the gate's current wording as the thing to follow.

Text 3: §2, the enum-value paragraph. It prescribed a hand-rolled enum error map dispatched on issue.input, citing HookBodyCapability (data/hook-body.zod.ts) and the object.managedBy: 'system' precedent. It now points at enumWithRetiredValues() in shared/retired-key.ts: declare the retired members and their prescriptions there, with no hand-written error map.

Kept: an enum-value narrowing is invisible to the four ratchets, and none of the three key routes applies to it.

Deleted: the two hand-rolled citations. Offered as the shape to copy, they now point at the wrong route. They also have another home: the helper's own docblock (retired-key.ts, "The gap this closes") names both as the hand-rolled precursors of the helper.

Measured before writing (origin/main at 79114850)

• The registry is generated. packages/spec/src/migrations/registry.ts:19-26 says the three tables are GENERATED from ./entries/, one file per entry, "never by typing between the markers. See ./entries/README.md." There are 237 files under entries/retired-keys/.

• The README owns the route. entries/README.md "Adding an entry" gives the file-name rule and the gen:migration-registry command. Its "Four rules" say that the next gen: run reverts a hand edit inside the markers, and that check:migration-registry (tsx scripts/build-migration-registry.ts --self-test --check) reports it before that.

• The helper has no production caller. HookBodyCapability (data/hook-body.zod.ts:48-60) and object.managedBy (data/object.zod.ts:1705-1710) are still hand-rolled z.enum(values, { error }) maps. enumWithRetiredValues() has 0 non-test call sites. git grep outside *.test.ts finds only its definition and docblock in retired-key.ts, and one test file uses it (shared/retired-key.test.ts).

• The helper covers every retirement the hand-rolled map covers. Both hand-rolled sites answer one exact retired string with a prescription and leave every other value to zod, which is exactly the helper's retired map. The helper's docblock names one case it does not cover: a retired member that WAS the default. A hand-rolled map cannot cover that case either, because the default must be a live member. The docblock routes that case to acceptRetiredDefaultResidue. So the kit says "don't hand-roll" with no exception, and stays true.

• No other sentence sends an entry into registry.ts. The census covered .claude/**, AGENTS.md and CLAUDE.md, grepping for migrations/registry, registry.ts, RETIRED_KEYS_BY_MAJOR, RETIRED_DEFS_BY_MAJOR, copy these lines and 粘贴, plus paste/append near registry. It found two registry.ts hits, both in the kit:

  • the text-1 line, fixed here;
  • the step-18 D3-chain item. That item is true and stays untouched: a MIGRATIONS_BY_MAJOR step's rationale and conversionIds are hand-written outside the generated markers (registry.ts:37-42; STEP18_RATIONALE at :5024).

  The kit's D3 semantic item already routes through entries/, and nothing outside the kit matched.

One route, said once in each place (with #20809)

At this head, #20809 has no branch or PR, so this lands first. Under the triage split the second to land quotes the first, so gate (b)'s printed remedy should name the same directory and command:

• directory: packages/spec/src/migrations/entries/retired-keys/, with the file name per entries/README.md;
• command: pnpm --filter @objectstack/spec gen:migration-registry.

#20809 remains open. This PR touches nothing under packages/spec/**.

Line budget

• Lines: 337 / 337 (ceiling 337, headroom 0). Text 1 went from 6 lines to 7, and text 3 from 4 lines to 3.
• Line width: the widest new line is 116 bytes, and every new non-table line is 120 bytes or less.
• Table rows: none touched. The widest is 324 bytes against the 326 pin.
• No re-wrap of untouched lines. Every deleted clause is false or has another home (see above).

Changeset

None: the diff publishes nothing. .claude/** ships in no package's files[]:

• 69 non-private package.json manifests, 0 of whose files[] names .claude;
• the root package is private;
• no .claude/ directory sits inside any package directory (git ls-files shows 50 tracked files under the root .claude/ only);
• positive control: @objectstack/spec's files[] reads dist, json-schema, liveness, prompts, llms.txt, README.md, src/**/*.zod.ts, CHANGELOG.md, api-surface, spec-changes.json.

The seat applies skip-changeset; this PR writes no label.

Gates (at head 21db5689)

node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (no paths) derived 18 commands, the same set as the dispatch. I ran all 18 plus three extras, check:pm-settings-deny-roster, check:pm-skill-id-lint and check:pm-governed-prose. Every one exited 0. dispatch-gates --ran with the recorded exit codes reported "18 derived famil(ies) accounted for — 18 run, 0 NOT-MEASURED (a DERIVED zero)".

• check:pm-skill-ratchet exit 0: "spec-property-retirement/SKILL.md is 337 lines (ceiling 337; headroom 0)"; "widest table row is 324 bytes (pin 326; headroom 2)".
• check:skill-frame-sync exit 0; check:doc-authoring exit 0; check:nul-bytes exit 0 ("no raw ASCII control bytes"); check:pm-skill-id-lint exit 0 ("34 file(s) clean").
• check:pm-governed-merges exit 0 · check:pm-governed-prose exit 0 · check:pm-settings-deny-roster exit 0.
• pnpm --filter @objectstack/lint run check:doc-formula-expressions exit 0, after a turbo run build --filter='@objectstack/lint...' (4 tasks, exit 0) under os-verify-lock.sh.
• check-ci-filter-parity · check-closing-keyword-parity (+ --self-test) · check-comment-mask-corpus · check-harness-current --self-test · check:agent-test-spelling · check:cross-package-test-inputs · check:driver-memory-census · check:gitlink-declared · check:refd-timer-probe · check:required-contexts · check:watch-hint-literal: each exit 0.

Acceptance notes

• A sibling of gate (b), same family as spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809. packages/spec/scripts/build-schemas.ts:2636-2640 is the def-removal gate. It prints the same "copy these lines in … under 18: [ … ]" remedy for RETIRED_DEFS_BY_MAJOR, whose table is generated from entries/retired-defs/. spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 lists only gate (b) (:1127-1128), so its author may want to take this sibling in the same edit. Not filed separately.
• Zero production callers for the helper. enumWithRetiredValues() still has no production call site, and its docblock @example shows HookBodyCapability built with it, which main does not do. Both hand-rolled precursors work as they are; this is an observation, not a defect. Carrier: none.
• Table-row pin headroom. The kit's table-row pin has headroom 2 (widest 324 against pin 326), and the ratchet prints a hint to lower it. That headroom predates this PR, and this PR touches no table row, so the pin is left as it is. Carrier: none.

---

Generated by Claude Code

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 21db56899a0085212b9627a60e4238510a973380
Local-runs: none

Read against: PR #20828 head 21db56899a0085212b9627a60e4238510a973380 (branch claude/issue-20803-retirement-kit-registry-entries, head repo = base repo, draft, 2 commits, 1 file, +8 / -8); origin/main at 72f8c3820154c69cdf9faa6f887dc544e4c4b823; merge-base 79114850f0f559dcb4fb432f5bcfc12947a03de6 (the dev's stated base). The kit blob at origin/main is 738187f1, the diff's own old side, so the PR's net diff against main is the one file, 8 / 8, exactly as GitHub lists it. Inputs: card #20803 body and its four comments (5906870733, 5907757995, 5907960440, 5908302870); #20809 body and pointer 5908348030 for context; PR body, file list, diff; PR thread: 0 comments. The template print (record-recognisers.mjs --template) is the brief's own prescribed read for this record's shape and evaluates nothing on the diff; every measurement below is wc / awk over git show output at the head or at main.

Check-runs on the head (filter=latest): 35 runs. 24 success, 11 skipped, 0 in any other conclusion. No job is outside success / skipped. Required contexts: Lint & Repo Gates success; TypeScript Type Check success; Test Core success (all six shards success); Dogfood Regression Gate success; Governed Surface Queue Guard success; Build Core skipped; Temporal Conformance (live PG + MySQL) skipped. The two skips are paths-filter skips on a .claude/**-only diff, not failures. Nothing is red.

① Derived judgments

Text 1: §3, the RETIRED_KEYS_BY_MAJOR item (head L182-188). Each statement checked at the code on origin/main 72f8c382; the registry header is byte-identical at the merge-base, and every line number the dev cited resolves at both.

• "add one file under packages/spec/src/migrations/entries/retired-keys/, naming per entries/README.md": RIGHT. entries/README.md "Adding an entry": one file named MAJOR.ID.ts with / and : spelled __; its table maps retired-keys/ to RETIRED_KEYS_BY_MAJOR[N]; 237 entry files exist there (same count at the merge-base). The kit does not restate the naming rule, as triage 5906870733 required.
• "run pnpm --filter @objectstack/spec gen:migration-registry, which generates it into registry.ts": RIGHT. packages/spec/package.json:296 gen:migration-registry = tsx scripts/build-migration-registry.ts; the generator splices src/migrations/registry.ts's marked regions (REGISTRY_PATH at :100, writeFileSync at :478).
• "never type there" (「⛔ 永不在那里手打」): RIGHT. registry.ts:19-26: the three tables are GENERATED, "never by typing between the markers"; README "Four rules": a hand edit inside the markers is reverted by the next gen: run and reported first by check:migration-registry (package.json:297 = --self-test --check; the --check branch at :464-475 exits 1 with "is stale — its generated regions do not match").
• Kept: the literal 'DEFKEY:NAME' value, spelled as authorable-surface/CATEGORY.json spells it minus [RETIRED], under the current major: RIGHT. build-schemas.ts:962 keys every property as ${defKey}:${name}; RETIRED_MARK = ' [RETIRED]' (:921) is stripped on read (:1016). Control pair: surface row api/AuthFeaturesConfig:magicLink [RETIRED] and entry file 17.api__AuthFeaturesConfig__magicLink.ts exporting 'api/AuthFeaturesConfig:magicLink'. The major is the filename prefix.
• Kept: gate (b) reads it by exact set membership, no inference, no radiation from a neighbour key: RIGHT. registeredRetiredKeys() folds the table into a Map (:942-952); gate (b) tests !registeredRetired.has(k) (:1117); :1304-1305 "There is no leaf-name matching left in this file (build-schemas.ts 检查 (c) 的「墓碑已满 2 个 major」证明仍用叶名匹配 —— 无关簇的登记可以替一次退休提前起算 #5898)".
• Kept: the (b2) warning, do not add the entry before the tombstone lands: RIGHT. (b2) at :1163-1185 fails on an entry naming a key the build still emits as LIVE.
• Kept: the "why a second table" note (build-schemas.ts 检查 (b) 用叶名匹配 conversion surface —— 无关簇的 .type 就能让一个 tombstone 冒充「已登记迁移」 #4659, endsWith('.' + name)) and the aging-clock note (build-schemas.ts 检查 (c) 的「墓碑已满 2 个 major」证明仍用叶名匹配 —— 无关簇的登记可以替一次退休提前起算 #5898, gate (c) reads the same table): RIGHT, :1100-1103, :1256-1260, :1303-1305.
• Deleted: 「在 packages/spec/src/migrations/registry.ts」 as the place to write the entry. FALSE at the code (a generated region); its replacement is the file route above. Not a lost rule.
• Deleted: 「门禁的失败输出会打印该粘贴的那一行」. Judged: at main today gate (b) does still print one 'KEY', line per unregistered tombstone under "copy these lines in" (:1127-1130), so the sentence is a true description of the gate's CURRENT wording; but what it describes is the paste-into-registry.ts route this card retires, and that wording is spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809's to change. The kit now names the file route and the command and quotes nothing of the gate's output, so it is true before spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 lands (the gate's remedy is wrong; the kit is not) and after (gate and kit name the same directory and command; pointer 5908348030 already carries the kit's wording to spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809, which lands second and quotes the first). The one true thing the sentence carried for an author, where the exact key spelling comes from, is kept in the item itself (the authorable-surface/ spelling rule). Not a lost rule; the gate's output is self-describing when it fires.
• Two lines of the old item survive byte-identical as diff context (the authorable-surface/ line and the (b2) tail line): confirmed.
• No other sentence in .claude/**, AGENTS.md or CLAUDE.md at the head sends an entry into registry.ts. Census (migrations/registry, registry.ts, RETIRED_KEYS_BY_MAJOR, RETIRED_DEFS_BY_MAJOR, copy these lines, 「粘贴」) hits only the kit: L182-184 (this text), L176 / L214 (conversions/registry.ts, a different, hand-written file), and L213, the step-18 D3-chain item, which is TRUE and untouched: a step's rationale and conversionIds are hand-written outside the generated regions (registry.ts:37-42; STEP18_RATIONALE at :5024). The three 「粘贴」 hits elsewhere are dispatch-word rules, unrelated.

Text 3: §2, the enum-value paragraph (head L141-143).

• "hang the prescription on the enum itself: enumWithRetiredValues() (shared/retired-key.ts) declares the retired members and their prescriptions": RIGHT. retired-key.ts:321-365: enumWithRetiredValues(values, retired) returns z.enum(values, { error }) whose map answers a retired member with its prescription and lets every other input fall through to zod's own message (:359-363).
• "no hand-written error map" (「⛔ 不再手写 error map」): RIGHT; the helper's contract covers both hand-rolled sites. HookBodyCapability (data/hook-body.zod.ts:48-60) answers exactly one string, 'crypto.hash', with CRYPTO_HASH_RETIRED; object.managedBy (data/object.zod.ts:1705-1711) answers exactly 'system' with MANAGED_BY_SYSTEM_RETIRED; both return undefined otherwise. The helper is that map generalised to N members (string-guarded hasOwnProperty lookup) plus three construction-time refusals a hand-rolled map lacks (an empty map, a "retired" member still in values, a blank prescription). It returns a plain ZodEnum, so .optional().describe(...) as managedBy uses it composes (docblock :285-288). Nothing either hand-rolled map does is outside the helper's reach.
• The disclaimed case (a retired member that WAS the default): the dev's claim that a hand-rolled map cannot reach it either is RIGHT. An enum error map runs only when a present value fails the enum: an omitted key materialises .default() without touching the map, and a released artifact carrying the retired default materialised is refused by ANY enum-level map, hand-rolled or helper. That population needs accept-and-strip ahead of the shape, which is acceptRetiredDefaultResidue (:189-218, a z.preprocess wrapper on the object, live at connector.zod.ts and permission.zod.ts), a different layer, not an error map. So "don't hand-roll" holds with no exception; the routing of the default case lives in the helper's docblock (:288-294) and was never in the kit, before or after.
• Kept: an enum-value narrowing is invisible to the four ratchets: still RIGHT at the code. authorable-surface/ records ${defKey}:${name} per property (build-schemas.ts:957-963); json-schema.manifest/ shards list def keys ("schemas": ["data/Address", …]); api-surface/ rows are name (kind); api-surface-signatures.json hashes checker.typeToString() of each defineX factory, which "prints a type REFERENCE and does not expand it … Value-level narrowing (an enum losing a member) is still ungated" (build-api-surface.ts:32-40, :177). The helper's docblock :300-303 restates the same.
• Kept: none of the three key routes applies to "def survives, one member leaves": RIGHT, helper docblock :257-259.
• Deleted: 「按 issue.input 分派」. Not false; that dispatch is now the helper's internals (:359-363). Its home is the helper. Not a lost rule for an author.
• Deleted: the two citations 「packages/spec/src/data/hook-body.zod.ts 的 HookBodyCapability,沿用 object.managedBy: 'system' 的先例」. As the shape to copy they now point at the wrong route; as precedents they still exist on main (both still hand-rolled) and are named as the hand-rolled precursors in the helper's "The gap this closes" (:255-268). Home kept; not lost.

Pins. Kit line count 337 at origin/main and 337 at the head (wc -l over git show); the ratchet rows are unchanged (scripts/pm/check-skill-line-ratchet.mjs:1377 ceiling 337, :2022 table-row pin 326) and the PR's file list is the kit alone. Added lines measured at 99 / 113 / 73 / 104 / 116 / 112 / 97 / 81 bytes: widest 116, all at or under MAX_LINE_BYTES = 120 (:2251). The only non-table line over 120 bytes at the head is L114 (311 bytes), a blockquote line (leading greater-than sign), present at main at the same line and structurally exempt (classifyLine 'blockquote', :2476). No table row touched; widest table row 324 bytes at L86 against pin 326. Text 1 went 6 to 7 lines, text 3 went 4 to 3: net 0, as the report says.

② Semver level

None: the diff publishes nothing. One file, .claude/skills/spec-property-retirement/SKILL.md, the only file in that directory; git grep for .claude over every package.json on main is empty and the root package is private. skip-changeset matches; Check Changeset is success on the head. Clause-②: no is right: no authorable key, export or config field changes.

③ Boundary flags

1. Tier naming. Triage unlock 5907757995 calls the card "a governed surface, Tier H". The register rules: scripts/pm/check-governed-merges.mjs:1094 row claude-tree, glob .claude/**, GOVERNED_TIER_S; AGENTS.md PD feat: Comprehensive CRM example demonstrating all ObjectStack protocol features #14 puts all of .claude/** in Tier S. The PR is Tier S; the dev's deviation note is right, and leaving another seat's comment alone was right. This record is the correction of record for the PR.
2. RETIRED_DEFS_BY_MAJOR paste remedy. Verified at build-schemas.ts:2636-2639: the def-removal gate prints the same "copy these lines in … under N: [ … ]" for a table generated from entries/retired-defs/; :789 names RETIRED_DEFS_BY_MAJOR (src/migrations/registry.ts) without saying to type there. packages/spec/**, domain:spec, outside this PR's claimed surface; carried to spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 as pointer 5908348030, which the seat posted. Escalated; nothing to do here.
3. enumWithRetiredValues() has 0 production call sites. Verified by git grep at origin/main 72f8c382: the definition and its docblock @example in retired-key.ts, plus one test file; HookBodyCapability and managedBy remain hand-rolled. This does not make the kit false: the kit prescribes the route for the NEXT narrowing, and the helper's docblock names both sites as the precursors it replaces. Converting them is a no-behaviour-change refactor, not a defect under PD chore: version packages #10; "noted, not filed" is the right disposition. Carrier: none.
4. Table-row pin headroom. Widest row 324 at L86 against pin 326 (:2022); the ratchet prints a lower-the-pin hint (:2592) and stays green. The row is untouched context in this diff and the headroom predates it; the claim limited ratchet edits to a lowered LINE ceiling, so leaving it is per the claim. Answered; a later edit to the file may lower it opportunistically.
5. Two required contexts read skipped. Build Core and Temporal Conformance (live PG + MySQL) are skipped on this head (paths filter, .claude/**-only diff). Not red; recorded so the landing seat reads the state as it is.
6. Commit trailers and extra gates. Both commits carry the model-free pair (Claude-Session: plus Co-authored-by: Claude), per AGENTS.md; the dev's deviation note about the harness reminder is right. The three extra gates run beyond the derivation are harmless.
7. Coordination with spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809. spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 still has no branch or PR; this PR lands first and spec: gate (b) of build-schemas.ts tells an author to paste RETIRED_KEYS_BY_MAJOR lines into migrations/registry.ts, which is generated from entries/ since #7297; two guidance-slot docblocks describe the old retirement route (spec half of #20803) #20809 quotes it (pointer 5908348030 already does). Nothing under packages/spec/** is touched here; the file list is one path.

Landing. Tier S: with this PASS on head 21db5689 and every check-run at success / skipped, the landing key holds; the owning seat lands it through the queue (ready first, arm second), and the post-merge audit is the compensating control.

Implemented-by: claude/issue-20803-retirement-kit-registry-entries
Reviewed-by: session_01KTZmMfzVzjNvyaLyQ8mHvg

VERDICT: PASS