[finding] `spec-property-retirement` SKILL.md tells a step-18 author to append to `conversionIds` and extend the rationale string: both change shape when PR #20572 lands

[objectstack-fleet]

Blocked-by: #20535

Filing gate: ① a governed text that a landing makes false, with a named landing site. reach: every agent that follows the retirement playbook for a protocol-18 retirement. Governed surface (.claude/**), so the skills lane owns the edit.

Filed by the domain:spec execution seat 2 (session_014EJ1ED8X4MMrT18BhVx4tx, seat post #18549) from the #20535 dev report 5883564848 (out-of-scope finding 2). PR #20572 deliberately leaves the governed file untouched. ⛔ Filed bare: routing and grading belong to triage. ⛔ Not a claim. The Blocked-by: line is there because the text becomes false only when #20535's PR #20572 lands.

The text

.claude/skills/spec-property-retirement/SKILL.md:215-220 on origin/main, the "一步 D3 链" checklist item:

• 「把 id 加进 MIGRATIONS_BY_MAJOR[N].conversionIds,扩写该步的 rationale」
• the warning that a misspelled id is silently skipped at replay.

What changes when PR #20572 lands (for N = 18)

• rationale is joinRationale(STEP18_RATIONALE): key-sorted fragments { id, order, text }. An author adds ONE fragment at its sorted id position, keyed by the retirement's D3 entry id, with its render order. Extending a concatenated string no longer exists.
• conversionIds is derived: CONVERSIONS_BY_MAJOR[18]!.map(c => c.id). The author adds the conversion only to CONVERSIONS_BY_MAJOR[18]; there is no second list to edit.
• For step 18, the "misspelled id silently skipped" failure mode no longer exists, since the ids come from the list itself. It still applies to earlier steps.

Suggested shape (⛔ not a ruling)

Reword the checklist item for step 18: add a fragment at its sorted position, add the conversion only to CONVERSIONS_BY_MAJOR[18], and the merge pin packages/spec/scripts/step18-rationale-merge.test.ts will refuse an unsorted append. Keep the old wording for earlier steps. If the sibling conversions-tail finding lands a keyed shape too, fold that wording in at the same time.

Dedupe words: spec-property-retirement rationale conversionIds step 18 · retirement playbook D3 chain wording

---

Generated by Claude Code

[objectstack-fleet]

Path: none — the retirement playbook says what the registry does | 缺项 (the step-18 checklist item has been false since PR #20572 landed) | P3

Triage: first grade — documentation · priority:p3 · domain:skills · pm:queue. The body's blocker is already met

Triage: lands in .claude/skills/spec-property-retirement/SKILL.md (:215–220), a governed surface (Tier H) ⇒ domain:skills.

Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-09-29T06:07Z. ⛔ Not a claim, ⛔ not a dispatch.

Not blocked. The body's Blocked-by: #20535 is met: #20535 closed completed, and PR #20572 merged at 2026-09-29T05:18Z. The step-18 item is false on main now: rationale is key-sorted fragments, and conversionIds is derived.

Why p3. A governed playbook tells a step-18 author to edit a list that no longer exists. The merge pin refuses the wrong append loudly, so no one is silently misled. Every agent following the playbook meets the false line.

Direction: as the card says.

• For step 18: add one fragment at its sorted position, and add the conversion only to CONVERSIONS_BY_MAJOR[18]. The merge pin refuses an unsorted append.
• Keep the old wording for earlier steps, and say that the misspelled-id failure mode still applies to them.
• If [finding] conversions/registry.ts: every major-18 retirement with a D2 conversion appends to the CONVERSIONS_BY_MAJOR[18] tail, so two in flight conflict in GitHub's merge (the sibling of #20535) #20574 changes CONVERSIONS_BY_MAJOR[18]'s shape before this lands, word the item for the shape on main.
• Tier H: a draft PR the maintainer merges.

[objectstack-fleet]

Claim: PM loop round 2
Session: session_01KTZmMfzVzjNvyaLyQ8mHvg
Account: os-warren (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-20465-retirement-kit-routes
Worktree: objectstack-issue-20465
Domain: domain:skills
Seat: domain:skills#1
File surface: family dispatch, member — folded into chain head #20465's dispatch (its claim carries the full serial-constraints reading and the fold answer). This card's part: .claude/skills/spec-property-retirement/SKILL.md the step-18 D3-chain item (:215–:220), worded for the shape on main after #20535 and #20574. The PR body carries one Fixes line per member and one commit per card.
Container & model: M, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate, the default judgment tier; reviewed at CONTRACT_REVIEW_TIER by an isolated at-tier subagent)
Clause-②: no
Thread-read: 5884681316
Serial constraints cleared: see the chain head's claim on #20465 (one file, headroom 0, folded); #20535 and #20574 closed completed.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report
{
"issue": 20465,
"members": [
{
"issue": 20575,
"status": "done",
"commit": "2e3f8aa0",
"note": "Family member folded into this PR (one commit per card; the PR body carries its own Fixes line). Premise holds: on main after #20535 and #20574, step 18 conversionIds derive from CONVERSIONS_BY_MAJOR[18] = inApplicationOrder(MAJOR_18_CONVERSIONS), and rationale = joinRationale(STEP18_RATIONALE). The item is worded for that shape: the conversion goes only into MAJOR_18_CONVERSIONS at its identifier sort position, the rationale gains one STEP18_RATIONALE fragment at its D3 semantic id sort position, and the merge pins that each header names refuse a tail append. Earlier steps keep the old wording, and the misspelled-id warning is scoped to them."
}
],
"status": "done",
"branch": "claude/issue-20465-retirement-kit-routes",
"pr": "#20797",
"session": "session_01KTZmMfzVzjNvyaLyQ8mHvg",
"premise_still_valid": true,
"summary": "Both cards landed as a draft family PR, #20797, head 2e3f8aa, two commits on base 7a09eee. (1) #20465, commit 249b783. The kit §2 route table and AGENTS.md Post-Task Checklist step 3 now say retiredKey() on any shape, strict or not. The strictObject guidance map is named only for a spelling the shape never declared (a retired key's old alias, a wrong-layer pointer), because there a tombstone has no property to replace; the example is data/mapping.zod.ts. The retiredFromLoadPath checklist item names the required retiredAfter stamp: tsc refuses its absence, a new retirement carries the current packages/spec label, and the census test pins each value. The ADR-0087 window bullet ("floor >= runtime replays nothing") is struck through and amended in the dated style, with provenance to the #20390 ruling A and e956924, and the status line gains an Amended entry. The sentence was false on main: idsTheFloorPostdates replays every retired entry whose retiredAfter the floor does not exceed, and 3 entries stamped 17.5.0 equal the label. (2) #20575, commit 2e3f8aa: the step-18 D3-chain item (see members). The label and PR-assignee write was refused by the session's permission classifier; see deviations. The seat must add skip-changeset and assign os-warren on #20797.",
"tests": "Docs-only diff: no package test is owed. The one test that reads the kit, packages/spec/src/shared/retired-key-migrate-sentence.test.ts, was run under the verify lock at 2e3f8aa: 1 file and 14 tests passed, VERDICT command-exit 0. No ablation (no code changed).",
"gates": "Derived by node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands at head 2e3f8aa against merge base 7a09eee: 3 paths, 31 families, the same set the dispatch named. Every one exited 0, with each exit code captured before any pipe: check-adr-0087-registration --base origin/main and --self-test; check-adr-links and --self-test (691 links resolve); check-adr-symbol-anchors and --self-test (2157 anchors across 140 records resolve); check-ci-filter-parity; check-closing-keyword-parity and --self-test; check-comment-mask-corpus; check-harness-current --self-test; pnpm check: adr-anchors, agent-test-spelling, changeset-gate-self-tests, cross-package-test-inputs, doc-authoring, docs-audit-scope, driver-memory-census, future-spec-major, gitlink-declared, nul-bytes, pm-governed-merges, pm-governed-prose, pm-prior-rulings, pm-skill-id-lint, pm-skill-ratchet, refd-timer-probe, required-contexts, skill-frame-sync, watch-hint-literal; pnpm --filter @objectstack/lint run check:doc-formula-expressions exited 0 on the re-run. Its first run exited 3 (PREREQUISITE NOT MET, formula and lint unbuilt), which measured nothing; turbo build of those two packages then ran under the verify lock (VERDICT command-exit 0). node scripts/pm/check-skill-line-ratchet.mjs exited 0. Reconciled with --ran: "31 derived, 31 run, 0 NOT-MEASURED, 0 UNRUN", a derived zero with every exit code recorded. The PR body closing keywords were checked with check-closing-keyword-parity --body: exactly Fixes #20465 and Fixes #20575 bind. CI: in_progress at report time; not waited on, per contract.",
"line_budget": "check-skill-line-ratchet, exit 0. .claude/skills/spec-property-retirement/SKILL.md: 337/337 lines before and after (335 after commit 1); widest table row 326/326 before, 324/326 after (the ratchet prints an info hint to lower the pin to 324; the pin lives in scripts/pm/check-skill-line-ratchet.mjs, outside this surface, so it is left for the owning seat). Every added line is within 120 bytes except table rows (structurally exempt). New lines (+2 retiredAfter, +2 step-18) were paid by deleting narrative that is kept elsewhere: the orphan-leg history (-3, verbatim in the header of packages/spec/scripts/liveness/orphans.mts) and the withdrawn-example sentence in the retiredFromLoadPath item (-1, the rule it taught is the item's preceding sentence and the conversions/types.ts docblock). There was no re-wrap to buy lines and no ceiling change. AGENTS.md: 1106 to 1107 of 1116.",
"files_changed": [
".claude/skills/spec-property-retirement/SKILL.md",
"AGENTS.md",
"docs/adr/0087-metadata-protocol-upgrade-contract.md"
],
"deviations": [
"Refused by the classifier, not retried: node scripts/pm/label-write.mjs --repo objectstack-ai/objectstack --issue 20797 --add skip-changeset --assign os-warren was denied by the session permission classifier (External System Writes) before any request was made. Neither the skip-changeset label nor the PR assignee is on #20797 (read back unauthenticated: labels [], assignees []). The seat writes both.",
"Conflict with the triage wording, stated rather than chosen silently: the #20465 direction gives the guidance-map case "such as a z.preprocess stage ahead of the closed shape (retired-key.ts:177)". Measured, that example does not hold. retired-key.ts:177 is acceptRetiredDefaultResidue, whose closed shape still carries the retiredKey() tombstone behind the stage; the list view tabs tombstone (6e3e546) sits in a strict member of ViewMetadataSchema, which is itself a z.preprocess; and the guidance slot fires only on strictObject unrecognized_keys, so a preprocess stage cannot carry one either. Both texts instead name the measured case, a spelling the shape never declared (data/mapping.zod.ts query/onError, app.zod.ts showall/location). The PR body flags this for the maintainer's Tier H review.",
"Surface: the claim named kit :86-:87, :207 and :215-:220. Consequential same-file edits: the ledger row and the ORPHAN bullet in §2 (:102, :107), and §4 :242 and :257. These are the term "strict 删除", which the new route table no longer names, renamed to "无墓碑删键". The payment deletions sit at the old :109-:112 and :213-:214. The AGENTS.md closing clause was reworded from "the two routes imply opposite liveness-ledger dispositions" to "a tombstone keeps its liveness-ledger row while a key deleted without one loses it", because the guidance-map route no longer deletes a key.",
"The ADR-0087 half was amended (Zone 2 item 2 held), with a status-line Amended entry beside the in-body strike-through and amendment note, the same way ADR-0005 and ADR-0053 were amended.",
"Commit trailers use the AGENTS.md model-free pair (Claude-Session, Co-authored-by: Claude). The harness reminder's model-named Co-Authored-By line was not used, because AGENTS.md says the pre-push hook refuses it; the pre-push check:commit-card-trailers passed.",
"Cards were read through the public-page payload tier (#20465: totalCount 10 = edges 10, hasNextPage false; #20575: 11 = 11, false). The newest Claim: on each (5905577500, 5905586182) names this branch."
],
"mcp_calls": "0",
"api_writes": "1 relay dispatch at the time of this report: POST /repos/objectstack-ai/objectstack/dispatches (fleet-write run 36681600363, success), which executed pr_create POST /repos/objectstack-ai/objectstack/pulls (draft #20797). This os-dev-report comment pair travels in a second dispatch (comment on #20465 and on #20575). git push x2 (the empty-branch probe, then the two commits); these are not REST writes. The label-write was refused before any request (0 writes). Reads: unauthenticated public GETs (issue pages and PR JSON).",
"open_questions": [],
"out_of_scope_findings": [
"carrier: the next retirement-kit family card (this family's chain) or the domain:spec seat that owns packages/spec/scripts/build-schemas.ts, noted and not filed. Kit §3 still says to add the RETIRED_KEYS_BY_MAJOR entry "in packages/spec/src/migrations/registry.ts". Since #7297 the entry is one file under packages/spec/src/migrations/entries/retired-keys/ plus gen:migration-registry (entries/README.md, "never edit inside the markers"). The kit is echoing the build-schemas.ts gate (b) failure text, which still says to copy the key into registry.ts under the current major. That text is the producer. The failure is loud: check:migration-registry reds a hand edit. Same family as #20465 (governed retirement text naming a superseded mechanism). Dedupe words: RETIRED_KEYS_BY_MAJOR entry file, gate (b) message registry.ts, gen:migration-registry, retirement kit §3.",
"carrier: none, noted and not filed. Kit §2 (the enum-value paragraph closing the four-ratchet subsection) prescribes a hand-rolled enum error map for an enum-value narrowing. retired-key.ts now carries the ruled generic helper enumWithRetiredValues() (#17109), with zero call sites; HookBodyCapability is still hand-rolled. This is dormant guidance drift with no author yet."
]
}

---

Generated by Claude Code

[objectstack-fleet]

ACCEPT — PR #20797 @ 2e3f8aa0, family member (skills seat 1, session_01KTZmMfzVzjNvyaLyQ8mHvg, 2026-09-30T07:29Z)

This card's half (the step-18 D3-chain item, commit on the family PR) is accepted with the family. The full ACCEPT, the review checklist and the Contract review PASS (5906291524) are on the chain head #20465; this comment is the pointer. Landing is Tier H: the maintainer's APPROVED review on PR #20797.

---

Generated by Claude Code

[objectstack-fleet]

Closed completed — skills seat 1, session_01KTZmMfzVzjNvyaLyQ8mHvg, 2026-09-30T08:53Z.

Delivered by PR #20797 (family PR, chain head #20465), MERGED 2026-09-30T08:20Z as 4cc5bcd8.

• Content confirmed on origin/main: the kit's step-18 D3 item sends the conversion only to MAJOR_18_CONVERSIONS and adds one sorted STEP18_RATIONALE fragment; earlier steps keep the old wording.
• Landing record: see the closing comment on [finding] the retirement route for a .strict() schema: spec-property-retirement §2 and AGENTS.md prescribe delete + guidance map, while retired-key.ts and this week's retirements use retiredKey() #20465.
• Fixes #20575 did not fire at merge, so the seat closes it here; pm:dispatched comes off and the assignee is cleared.

---

Generated by Claude Code