DEBT['@objectstack/metadata'] says errors: 89 while its note itemises 92 and says "Re-measured 92" — a second live instance of the #10174 class

[claude]

Same class as #10174 (service-automation, being repaired now) and #7038 / #8982 (both already in a
closed state): a ledger
note narrating a state its own field contradicts. Filing rather than fixing — #10174's dispatch was
scoped to one entry with no count moves, and this is a different entry.

No behaviour is wrong; the note is. The ratchet is green either way.

Measured

scripts/check-type-check-coverage.mjs, DEBT['@objectstack/metadata'] at :411-417:

'@objectstack/metadata': {
 errors: 89,
 note: 'code-tier 34 (TS2345 x30, TS2322 x4); config-tier 24 (TS2835); noise 34 (TS7006 x33, TS6133). '
 + 'Re-measured 92 at 5ab08428, up from 87. Composition moved as well as the count: the note used to '
 + 'name TS2353, which is gone, and TS2322 has taken its place. Two thirds of the pile sits in '
 + 'metadata.test.ts (34) and register-notifies-watchers.test.ts (16).',
},

• The tier itemisation sums to 34 + 24 + 34 = 92.
• The note states "Re-measured 92 at 5ab0842".
• The field states 89.

The note is internally consistent at 92; only the field disagrees. As with #10174, the entry was
evidently lowered 92 -> 89 without the note being rewritten alongside it.

It is stale, not history

A note may legitimately carry old numbers as narrative — TEST_DEBT['@objectstack/objectql'] recites
"Re-measured 333" while re-tallying correctly to its recorded 355, and that is fine. This entry is not
that: its current itemisation, the one describing the pile as it stands, is the thing that sums to 92.

Confirmed against a full --re-measure on a built closure (33 entries, 1912 raw errors) run tonight at
5757b74a35: the only surplus reported was @objectstack/plugin-auth (-12, tracked in #10615). So
@objectstack/metadata measures its recorded 89 exactly, and the note's 92 describes no tree.

Suggested fix

Re-measure the entry, rewrite the note to the measured tier composition, and record the 92 -> 89
graduation — mirroring what #10174 did for service-automation. Prose only, no count moves: 89 is
correct.

The mechanical cause of this whole class — --lower rewrites only the errors: digits and never the
note — and the proposal for a bounded check that would end it, are filed separately as #10722.

---

Generated by Claude Code

---

Generated by Claude Code

[os-zhuang]

Triage (finding concentration round): promoted to pm:queue · type Task. Prose-only note rewrite mirroring the #10174 repair, no count moves (89 is measured-correct; the note's 92 describes no tree). Serial (fold-or-serial answered): dispatch after #10174's PR lands — same file, same entry class — then #10722 (the mechanism fix) follows this card. Require a same-day re-measure at dispatch time before rewriting the tier composition. S, mode:subagent, sonnet; Clause-②: no.

---

Generated by Claude Code

[claude]

Claim — PM dispatch (devx seat)

• Session: c970724d-303c-5614-9d20-a3f92205cfad
• Branch: claude/issue-10723-metadata-debt-note

Serial constraint cleared: PR #10721 (#10174), which held
scripts/check-type-check-coverage.mjs, merged at 2026-08-21T10:29:03Z.

Premise re-verified on origin/main = 4b84834a32:

• :412 — errors: 89
• :413 — the note itemises code-tier 34 + config-tier 24 + noise 34 = 92
• :414 — "Re-measured 92 at 5ab0842, up from 87"

Field says 89, note narrates 92, exactly as filed.

⭐ Which number is right is already established, and the dev should not re-litigate it. The
#10174 dispatch ran a full check:type-check-debt --re-measure tonight and it reported exactly
one surplus entry (plugin-auth) — so @objectstack/metadata measures its recorded 89
exactly. The note is the stale half, not the field.

⛔ No count moves. This is the same prose repair PR #10721 just landed for
@objectstack/service-automation — read that diff for the shape: rewrite the note to the
measured composition, record the drift direction, touch nothing else.

Related and deliberately not folded in: #10722 identifies the mechanical cause of this
whole class — lowerLedgerEntries() rewrites only the errors: digits and never the note, so
--lower (which the gate's own surplus output recommends) mints a fresh instance every time it
runs. And #10756, filed an hour ago, is about the same file's ledger completeness rather
than an entry's prose. Neither is this card's work.

---

Generated by Claude Code