[rebuild of #19384] check-widening-tells T2 routes any bare string element to "a new member of a closed set" without measuring whether a closed set encloses it

[os-steve]

⭐ REBUILD of card #19384, whose original is unreachable. Filed by the domain:spec seat 4 (session_01AmH9bKvGoLjiY86Q4Z3og2, seat post #18917) on the maintainer's instruction 「os-sam 被封号，19384 你要重建」, 2026-09-21.

⛔ The original card is not deleted and ⛔ nothing here overrules it — it is unreadable. Measured at this act, exit codes captured before any pipe:

path	reading
GET /repos/objectstack-ai/objectstack/issues/19384	404 Not Found
PATCH …/issues/19384	404 — so it cannot be edited or closed either
…/issues/19384/comments	200 — 12 comments still resolve
…/issues/19384/timeline	200

⇒ the issue's sub-resources resolve while the issue node does not. The common factor across every card in this lane showing the same shape — #19384, #19474, #19377, #19389 — is that each was filed by os-sam, the banned account. ⛔ This is the ban, ⛔ not a transient fault: an earlier reading by this seat called it transient and that reading is withdrawn.

History that still reads, and is the authority over this rebuild: the original card's 12 comments, and PR #19438 (claude/issue-19384-t2-closed-set-opener, draft, 1 file, +1282/−28, head c222a58309), whose body carries the fullest statement of the defect.

The defect, as PR #19438 states it

check-widening-tells.mjs's T2 doctrine names four closed-set forms — z.enum([…]), z.union([…]), z.discriminatedUnion(…), and a CORE_PLUGIN_TYPES-shaped as const array. The classifier required none of them: a lone quoted string, or a bare …Schema, arm, anywhere on the contract-source surface was routed to T2, and the row then rendered 「a new member of a closed set … the accept set gains a value」 over it.

⭐ That is a fact the instrument never measured. It is ⛔ not covered by the file's own 「a tell, never a proof」 allowance, which excuses a tell that is right about its own shape and silent about direction — this one is wrong about its own shape.

⚠️ The card's line anchors drifted three times, each re-read from source rather than adjusted by arithmetic: BARE_STRING_ELEMENT :1962 → :1957 → :2111; the memberTellKind arm routing to T2 :3373 → :3368 → :3644. The drift is d9282a4bd7 (#19153, +337/−27 in that file), which landed the T1 leg.

Disposition — ruled, ⛔ not this seat's judgment

The director seat's pointer on the original card (5755286648, 2026-09-21T04:13Z) records ruling #208 on #19491, maintainer 「19491 接受你的建议，并立刻派发处理相关任务。」:

ruling #208 on #19491 … makes check-widening-tells report-only and stops its development; the other four false-positive cards (#19341, #19221, #19440, #17926) are closed not_planned in this stroke. This card and PR #19438 carry the spec seat's claim and assignee, so this seat does not touch them: the spec seat closes both under the same ruling … or says on this card why not.

⇒ this rebuild exists so that disposition has a carrier it can actually be written on. This card is closed not_planned in the same stroke as its creation, and PR #19438 is closed with it — the work stopped at round 8 and that is already on this seat's record.

⛔ Not closed because the finding is wrong. The defect is real and was measured across eight rounds; what the ruling settles is that a report-only instrument does not earn dev time for being wrong. If check-widening-tells is ever made to hold a hard gate again, this card is the record of what T2 does not measure.

查重词

check-widening-tells T2 closed set · BARE_STRING_ELEMENT bare string element · memberTellKind T2 arm · closedSetMembership enclosing delimiter · rebuild of 19384 os-sam ban

---

Generated by Claude Code