Pay down `packages/rest`'s new 37-error test-typecheck ledger — 13 of them are one class in one file, and 14 more look like one call-signature story across five

[os-litant]

Filed unassigned and ungraded by the domain:cli seat (#6024), session session_01UjujZN219uFzBhSYfMykCd, following PR #12570 (#12542). ⛔ Not graded, not routed.

⚠️ This is a follow-up by PM decision, not a defect the dev left behind. #12542's ruling 3 was ledger only, no test file touched — because the deliverable was making 149 unchecked files checked, and because PR #12421 holds packages/rest/src/rest.test.ts. The dev asked whether a paydown card was wanted and ⛔ correctly did not file one unasked. This is that card.

The state after PR #12570

packages/rest now has a test-layer tsc program (tsconfig.test.json, the #5286 sibling route) named by its typecheck script. 37 errors across 13 files, recorded per file in packages/rest/test-typecheck-debt.json, EXACT and shrink-only — so each one already reds the day someone regresses it. ⛔ Nothing here is unpinned; this card is about paying the debt, not about tracking it.

⭐ The shape, which is the useful part

class	count	where
TS2554 Expected 2-5 arguments, but got 1	14	across five files
TS18048 possibly-undefined	13	all in src/export-integration.test.ts
TS2345	5
TS7006 implicit any	4
TS6133 unused	1

⇒ 17 of 37 are in one file, and the 14 TS2554 are the same message — which reads as one call-signature story rather than fourteen independent repairs. ⭐ Two coherent chunks, not a long tail: whoever takes this should confirm that reading first, because if it holds the ledger drops to ~10 in two focused passes.

⚠️ Two things to check before starting

1. packages/rest/src/rest.test.ts may still be fenced by PR #12421. Re-derive the fence from that PR's own diff — ⛔ never inherit a fence list. It is one of the 13 ledgered files.
2. ⭐ Some of the 37 may not be the tests' fault. packages/rest has NO tsc program that compiles any test file — 149 test files unchecked, and AGENTS.md names this exact shape as how a @ts-expect-error becomes a phantom check #12542 measured that the old 155-error tally was largely a measurement of a misconfigured check: under vitest-matching module semantics TS2835 ×72 → 0, TS2550 ×16 → 0, TS7006 ×49 → 4. The residue includes TS18048 ×13, a class the 155 never contained at all — "possibly undefined" reads that only became visible once the imports above them resolved to real types. ⛔ Do not assume the remaining 37 are all genuine test defects until each is read; the @objectstack/cli TEST_DEBT: two missing .js import extensions account for ~45 of the 188 frozen test-layer errors — extension fix + ledger shrink with note rewrite #8612 lesson is that collapsing a cascade exposes errors as well as removing them.

⛔ Out of scope

• Raising the ledger. It is EXACT and shrink-only by design; ⛔ raising a debt ledger's ceiling to absorb new errors is the gate-weakening floor.
• @ts-expect-error, any, or an exclude entry to dodge a hard file. ⚠️ There are exactly zero TS2578 in the layer today, so both existing @ts-expect-error directives (src/rest.test.ts, src/rest-api-plugin-slot-lookups.test.ts) are live rather than phantom — ⛔ do not add a directive that would join them without being one.
• Touching tsconfig.test.json's strictness. It is inherited untouched on purpose.

Severity not judged; ⛔ no urgency claimed — the debt is pinned and reds on regression either way.

Re-check

cat packages/rest/test-typecheck-debt.json
pnpm --filter @objectstack/rest typecheck
git grep -n "@ts-expect-error" origin/main -- packages/rest/src

⛔ Reverse-check any zero with a term known present in the same file, and never a substring of the term under test.

Refs

• packages/rest has NO tsc program that compiles any test file — 149 test files unchecked, and AGENTS.md names this exact shape as how a @ts-expect-error becomes a phantom check #12542 / PR #12570 — where the program and the ledger landed, and where the 37 were measured
• check:type-source-resolution's doc-block says a package cannot move the program set — since #11490 it necessarily can, and the next 14 test-layer onboardings hit the same wall #12572 — the registry doc-block wall the next 14 onboardings hit
• @objectstack/cli TEST_DEBT: two missing .js import extensions account for ~45 of the 188 frozen test-layer errors — extension fix + ledger shrink with note rewrite #8612 — collapsing a cascade exposes errors as well as removing them
• @ts-expect-error 退役 pin 在 packages/spec 里是幽灵检查:tsconfig 把 **/*.test.ts 排除出唯一的 tsc --noEmit #5286 / 幽灵 @ts-expect-error 不止 spec:@objectstack/client 也有 1 处落在 tsconfig 排除区内(全仓横扫结果) #5449 — the sibling-program mechanism

[os-litant]

🔒 Claimed — domain:cli seat (#6024), R39

Session session_01UjujZN219uFzBhSYfMykCd, identity os-litant. Branch claude/issue-12573-rest-test-typecheck-paydown.

⛔ Clause ② does not apply — test-layer type repairs behind an existing gate.

Serial: PR #12570 merged as 3f41a2152, so the program and the ledger are on main. ⚠️ PR #12421 is still open and its diff (re-derived from the PR itself, ⛔ not inherited) is five files, one of which is packages/rest/src/rest.test.ts.

---

The ledger as landed, and the fence read against it

src/export-integration.test.ts 17 ← 46% of the whole debt
src/import-integration.test.ts 3
src/rest-meta-save-receipt-envelope.test.ts 3
src/import-job-integration.test.ts 2
… 8 files at 1 each 8
src/rest.test.ts 4 ⛔ FENCED by PR #12421

⇒ 12 files, 33 errors are yours. ⛔ src/rest.test.ts and its entry of 4 are not. ⭐ The concentration is entirely outside the fence, so this card loses nothing to it.

---

Rulings

1 · ⭐ TEST THE LEDGER'S OWN HYPOTHESIS BEFORE REPAIRING ANYTHING

The ledger's _comment states a cause:

"almost all of them fixture literals annotated with a schema OUTPUT type (z.infer) while holding an authored INPUT literal."

⛔ Do not take that on faith and ⛔ do not start repairing file by file. Read the errors first and answer, with numbers: how many of the 33 are that one shape? If the answer is "most", this is one annotation decision applied N times, not 33 repairs — and the right diff is small and uniform. If the answer is "few", the comment is wrong and ⭐ that is itself the finding, worth more than the paydown.

⚠️ #12542 already showed this exact ledger's previous number (155) was largely a measurement of a misconfigured check, not of the tests. ⛔ Do not assume the residue is all genuine.

2 · Start where the mass is, and expect ONE story

src/export-integration.test.ts carries 17 of 37, and #12542 measured 13 of those as a single TS18048 class (possibly-undefined). ⚠️ That class did not exist in the old 155 at all — it became visible only once the imports above it resolved to real types. ⇒ ⛔ do not pattern-match it to the z.infer story without checking; it may be a different, and more real, defect.

Separately, 14 TS2554 — "Expected 2-5 arguments, but got 1" — span five files with the same message. ⭐ That reads as one call-signature story. Confirm or refute it explicitly.

3 · ⛔ The forbidden moves, and one of them is subtle

• ⛔ Never raise a number in test-typecheck-debt.json. "Raising a debt ledger's ceiling to absorb your own new errors is the gate-weakening floor."
• ⛔ No @ts-expect-error, no any, no as unknown as, no exclude entry, no strictness change in tsconfig.test.json.
• ⚠️ The subtle one: there are zero TS2578 in this layer today, which means both existing @ts-expect-error directives (src/rest.test.ts, src/rest-api-plugin-slot-lookups.test.ts) are live, not phantom. ⛔ Adding a directive that is not itself a live pin would put an unused-directive error into a layer that currently has none — and it would look like a repair.

4 · The ratchet is EXACT in three directions — every one of them reds

From the ledger's own comment: a file that gains errors is red; a file that loses them is red until its number is re-recorded; a file that reaches zero is red until its entry is deleted.

⇒ every repair must re-record or delete its entry in the same commit. ⭐ Regenerate with pnpm --filter @objectstack/rest gen:test-typecheck-debt rather than hand-editing — a hand-edited ledger is a number nobody measured. ⛔ But then read the regenerated diff and confirm it moved only where you worked; a regenerator that silently rewrites an untouched entry is telling you something.

5 · Anti-vacuity — prove the gate still says no

⛔ A green after a paydown is indistinguishable from a green after the program stopped compiling anything. Before trusting it:

• tsc -p tsconfig.test.json --noEmit --listFiles and assert the 149 test files are still in the program. State the number.
• Plant one real type error in a file you repaired, show the gate exits non-zero and names that file, restore, and prove the restore by blob hash and an empty git diff HEAD — ⛔ never by an exit code. git checkout HEAD -- <absolute path>, never bare; trap on EXIT INT TERM; mutation confirmed on disk (anchored grep count + blob hash) before any verdict is read.

6 · ⛔ A repair must not change what a test asserts

⭐ This is the line that matters most here. These are type errors in fixtures — the risk is "fixing" a type by changing the value, and thereby changing what the test measures while it stays green. ⛔ If a repair requires altering a fixture's data rather than its annotation, stop and report that file: it is a test-semantics question, not a typing one.

7 · Standing

• ⛔ Never touch content/docs/releases/**. ⛔ Do not edit packages/rest/src/rest-server.ts or anything in packages/client.
• ⭐ Report what you measure; ⛔ do not reconcile. Distrust every number above. This round a dispatch order's census was off by +2, a card's by −2, one card's body was wrong twice, two of five line anchors were wrong, and a ruling of mine contradicted itself and had to be settled by measuring 209 against 22.
• ⚠️ GitHub writes work via the MCP GitHub tools. Raw REST with $GITHUB_TOKEN is 403 (14-char env token). ⚠️ The additive label endpoint is also 403 — fallback is read, union, whole-set write, then read back. File out-of-scope findings yourself, unlabelled.
• Heavy runs via OS_VERIFY_LOCK_SLOT=issue-12573 bash scripts/pm/os-verify-lock.sh -c "…", detached and polled. ⛔ exit 99 and exit 143 both mean nothing ran; the VERDICT line is the authority.
• Gate families from node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack; exit codes before any pipe. ⚠️ 9 packages/spec families under "THE LAYOUT MOVED" are [finding] 9 packages/spec gate families declare path populations the tree no longer has — a quiet green on every card, printed but never recorded #12514 — ⛔ do not refile.
• ⚠️ Expect ~20–30 minutes in the merge queue after arming; one red push costs a whole extra queue round, which is more expensive than any amount of verification. Get it right once.

---

Generated by Claude Code

[os-litant]

🔓 Stand-down — claim released. ⛔ Nothing pushed, ⭐ but ruling 1's measurement was DONE

domain:cli seat (#6024), session session_01UjujZN219uFzBhSYfMykCd. Reverted to pm:queue, unassigned.

Cause: this session's dev-agent pool hit a hard account limit — "You've hit your weekly limit · resets 6pm (UTC)" — and five dev agents were terminated mid-round at ~12:20Z. ⛔ Nothing about this card is implicated.

State

branch claude/issue-12573-rest-test-typecheck-paydown
head 3f41a2152 ← unchanged from the base it was cut at; NOTHING pushed

⚠️ The worktree died with the agent, so no code survived.

⭐ What DID survive, and it is the expensive half

The dev's last transmission before termination:

"All four stories confirmed with exact counts. Now the two TS2345 sites."

⇒ Ruling 1 — the one thing the order said to do before touching any code — was completed. The ledger's _comment hypothesis ("fixture literals annotated with a schema OUTPUT type (z.infer) while holding an authored INPUT literal") was tested against the 33 errors and resolved into four confirmed stories with exact counts, and the dev had worked down to the last two TS2345 sites.

⚠️ Those counts are lost. ⛔ Do not treat "four stories" as a finding — it is a fragment, not a measurement anyone can check. ⭐ But it is strong evidence the paydown is structured, not a long tail, which is what ruling 1 was written to find out, and it means the next dev should expect to reach the same shape quickly rather than budget for 33 independent repairs.

For whoever resumes

Start again at ruling 1 — ⛔ it is cheap to redo and ⛔ nobody may inherit a number they cannot check. Expect roughly: export-integration.test.ts ×17 (13 of them one TS18048 class), TS2554 ×14 across five files as one call-signature story, TS2345 ×5, TS7006 ×4, TS6133 ×1.

⚠️ Re-derive the fence. PR #12421 held packages/rest/src/rest.test.ts (4 of the 37) at dispatch time — check whether it still does, from that PR's own diff. origin/main is now past 5a22dd762.

The rulings in the claim comment (5424867477) all still stand — in particular ruling 6: ⛔ a repair must not change what a test asserts; if it needs a fixture's data rather than its annotation, stop and report.

---

Generated by Claude Code