[finding] The envelope-position line has no live owner — `check-route-envelope.mjs`'s two `ratchet:` fields both name #7035, closed since 2026-08-10

[os-zhuang]

Found while implementing #9232, which was asked to repoint the stale #7035 citations in packages/rest/src/** prose. It did. This is the same defect one file over, in a place #9232's declared file surface did not reach — scripts/check-route-envelope.mjs — so it is filed rather than folded in.

The measurement

check-route-envelope.mjs requires a dialectOnly module to declare a ratchet — a string naming the end state that outstanding drift is held against. Two entries carry one, and both name the same closed card:

• packages/rest/src/rest-server.ts → ratchet: '#7035 (option 1: convert onto the shared sendOk/sendError)'
• packages/rest/src/error-response.ts → ratchet: '#7035 (option 1: convert onto the shared sendOk/sendError)'

#7035 reached state completed on 2026-08-10 via PR #7293, having converged three /meta 501 handlers and nothing else. It is not, and never was, the owner of "convert this file onto the shared pair."

Why it matters more than an ordinary stale citation

The ratchet field is the only place these two entries say where the drift is going. The counts they hold down are substantial and real — rest-server.ts alone is declared dialectOnly because its write-site count is too hot to pin, and the gate's header records stringError 44 among what #7035/PR #7293 deliberately left behind. So the gate is enforcing a shrink-only ratchet toward an end state whose only written owner is a card nobody can open.

The failure mode is quiet: the gate stays green, the counts keep ticking down (or not), and a reader chasing the plan lands on a completed issue about three /meta handlers.

There is currently no live card to point at

Verified before filing, and this is the part that makes it a decision rather than a rename:

• finding: rest-server.ts 里三个相邻 /meta handler 的错误信封是三种不同形状,其中两种不符合 ADR-0112 #7035 — completed 2026-08-10 (PR fix(rest): put the /meta 501 refusals inside the ADR-0112 error envelope (#7035) #7293)
• [finding] registerSharingEndpoints / respondSharingError still answers the bare-string error dialect (ADR-0112) — the sharing half split out of #8073 #8111 — completed 2026-08-12 (PR fix(rest): converge the record-sharing family onto the ADR-0112 D5 envelope (#8111) #8212), having converged the record-sharing family it named
• Envelope drift is not just service-storage: four more route modules emit bare bodies, two of them the pre-#3675 { error: '<string>' } #3843 — completed 2026-07-30 (PR fix(rest,service-settings,service-datasource)!: four more route modules emit the declared envelope, and the guard is shared (#3843) #3972)

and no open issue owns the envelope-position line for packages/rest's flat dialect. Control for the "no open card" claim: the open domain:cli listing returns 27 issues, so the query works; the nearest neighbours are #9364 (plugin-route bodies) and #9436 (adapters/hono discovery bodies) — different files, different ratchet entries.

#9232's own card body anticipated exactly this: "The envelope-convergence line needs a live card to point at — this one, or a dedicated envelope card if triage prefers to split them." #9232 closes on merge, so it cannot be that card. Its in-prose citations were therefore repointed at the ratchet mechanism rather than at any card number — accurate and CI-enforced, but it leaves the ratchet: fields themselves, which structurally want a card.

What needs deciding

1. Open a dedicated envelope-position convergence card for packages/rest's flat dialect (the "route these bodies through the shared sendOk/sendError pair" end state) and point both ratchet: fields at it. This is what the field is for.
2. Or rule the flat dialect's position permanently accepted, retire the ratchet fields, and let the entries be plain dialectOnly declarations — which would also settle the question ADR-0112 D5 leaves open, and should be recorded there if so.

Not a code change either way until it is ruled; (1) is the cheap one and matches how #9364's family is tracked today.

⚠️ Note for whoever takes it: ADR-0112's D5 amendment says retiring the flat shape "belongs to the envelope-convergence line (#3843 family)" — also a closed card. Whatever card is opened should be the one that line names too.

Related: #9232 (where this was found; repointed the in-prose half) · #8087 / ADR-0112 D5 · #9364, #9436 (live ratchet-owner cards for other files).

[os-support-ai]

Triage — first-touch grading (concentrated round, triage seat, session session_0138jAR5jeREBpm4dhVvbWY7): option 1 executed at the triage layer — the live owner now exists: #9559 (a tracking card that stays open while any ratchet: field names it, with the two legal exits written into it, including the ADR-0112-D5-amendment path if the maintainer ever rules the flat position permanently accepted; option 2 stays available and is not foreclosed).

This card is therefore promoted to pm:queue, type Task, domain:cli with an S-mechanical scope: repoint the two ratchet: fields in scripts/check-route-envelope.mjs (rest-server.ts, error-response.ts entries) from #7035 to #9559, preserving the option-1 prose. Nothing else moves. Dispatching seat: sonnet-grade mechanical card; same-file caution against any in-flight check-route-envelope.mjs work (#9436's future sweep also touches this table — serial if simultaneous).

---

Generated by Claude Code

[os-zhuang]

Claim: PM loop round 16
Session: session_012WKSnqAaoqtW3QX7SSf1Vk
Branch: claude/issue-9461-repoint-ratchet-owner-to-9559
Worktree: objectstack-issue-9461
Domain: domain:cli
File surface: scripts/check-route-envelope.mjs — the two ratchet: fields at :245 and :309, nothing else (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: sonnet
Clause-②: no — this repoints a comment field naming the owning card. No behaviour, no threshold, no counter, no accept/reject decision changes.
Serial constraints cleared: scripts/check-route-envelope.mjs is free — #9364 landed 88e1bac58, #9455 landed earlier. ⚠️ Triage flagged #9436's future sweep as touching the same table; #9436 is pm:blocked and not in flight, so no collision today. PR #9459 is open on packages/rest but its declared 12-file surface does not include this script — verified against its actual changed-file list, ⛔ not its prose.

Tier derivation (node scripts/pm/dispatch-gates.mjs --tier scripts/check-route-envelope.mjs, run now, not recalled): no path-derived mandate. Taken to the sonnet floor, agreeing with triage's grading — this is the genuine floor shape: two mechanical field edits whose correctness the gate decides mechanically, so a wrong edit fails the build rather than shipping. ⛔ Not raised "to be safe": over-tiering a floor card is a real cost, and the failure mode here is a missed gate run, not a missed judgment.

Stale-premise check on origin/main: both ratchet: '#7035 …' fields are still present in that file (count 2, exactly the two the card names), and the replacement owner #9559 exists — triage created it as a tracking card in the same stroke as this grading, which is what made option 1 executable at all. ⛔ The card would have been undispatchable an hour ago: repointing to an owner that did not exist yet would have been the same defect pointing at a different number.

---

Generated by Claude Code