check-half-states self-test: route every nullable predicate row through a wrapper, and delete the 14 literal stray spaces by claude[bot] · Pull Request #12390 · objectstack-ai/objectstack

claude · 2026-08-26T00:02:33Z

Two disjoint cards, folded into one PR by the skills-lane PM: a self-test
robustness fix in scripts/pm/check-half-states.mjs, and a typography pass over
three governed corpus files. The file sets do not overlap.

⛔ Draft, governed surface. The diff touches .claude/**, so under Prime
Directive 14 this is for the maintainer to land by hand: never flipped ready,
never queued, no auto-merge armed.

Card 1 — self-test nullable-row robustness

What the defect is

The self-test's message assertions read predicate(...).includes(needle) on
predicates that are three-valued by design — null when the card is clean,
a string when the row fires. The moment a change makes such a predicate go
clean, that line throws:

TypeError: Cannot read properties of null (reading 'includes')

The throw happens while evaluating t()'s arguments, before t() runs, so
no harness-level catch can turn it into a case. And because t() only pushes
onto a cases array that is printed in one loop at the very end, a single throw
means the suite prints nothing at all — not a partial run.

⚠️ The population is 253, not 111 — the card's table measured one of two shapes

The card counted direct call expressions (pred(...).includes(). Re-derived
mechanically on the base commit, that population reproduces exactly: 111
bare sites across the 18 tabled predicates. But it is not the whole shape.

population	what it looks like	sites
A — direct call expression	`h9OnHoldNoRestartWhen(hold()).includes('x')`	145
B — variable-mediated	`const h9NoLine = h9OnHoldNoRestartWhen(...)` … then `h9NoLine.includes('x')`	108
	total unsafe receivers	253

Population A's 145 = the card's 111 bare, plus 4 already written ?.includes
(safe from the throw, but reporting undefined rather than a clean false),
plus 30 reached through two local helpers the table did not name (halves
wrapping H8, dead27 wrapping H27).

Population B is invisible to any call-expression scan, and it spans four
predicates the card never mentions — h20, h21, h23, h28. It is also what
actually mattered: with population A alone fixed, the ablation below still
aborted, at h9NoLine.includes('cannot parse').

The fix

Per the triage direction: the per-predicate wrapper pattern PR #12272
established for H34, const h34row = (...args) => String(fn(...args) ?? ''),
generalized. Not a t() overload — the throw precedes t(), so the harness
cannot catch it.

18 wrappers (h4row … h33row) declared in one block at the top of
selfTest, plus halvesRow and dead27Row beside their own local helpers.
145 call sites switched.
Population B, two shapes chosen per variable: 19 variables carrying no
nullability assertion had their declaration wrapped in place (zero
call-site churn); the 10 that are asserted with typeof X / X === null kept
the raw const and gained an XRow sibling (56 sites switched) — stringifying
those in place would have made their own assertions trivially true.

⚠️The predicates are unchanged, and are still asserted three-valued
directly. Every typeof pred(...) === 'string' and pred(...) === null case
deliberately does not go through a wrapper.

An audit deriving the nullable set from the source (functions containing
return null) rather than from a hand-kept list now reports TOTAL A = 0,
TOTAL B = 0.

Acceptance (a) — full self-test green, case count unchanged

node scripts/pm/check-half-states.mjs --self-test → exit 0:

✓ check-half-states self-test: 1380 cases pass.

Stronger than "green": the run's entire stdout is byte-identical to the
pre-change baseline — same 1380 case names, same verdicts, same order. No case
changed its expected verdict, because no case changed its observable value.

Acceptance (b) — the ablation

⚠️ The card's own leg A no longer reproduces, and that is a real result

The card quotes a TypeError abort from widening CLAIM_COMMENT_MARKER to accept
U+2014. Re-run here on both the pristine base and the fix, that mutation now
behaves identically on both sides:

LEG before (base 7986d973f): exit 1 · 1381 cases printed · 12 named ✗ · TypeError lines: 0
LEG after (this branch) : exit 1 · 1381 cases printed · 12 named ✗ · TypeError lines: 0

All 12 failures sit in the H33/H34 blocks — the rows PR #12272 had already
wrapped. The abort the card recorded was observed before that wrapper
landed; once it did, this leg stopped aborting. So the card's before/after shape
could not be demonstrated with this mutation, and forcing it would have produced
a fabricated contrast.

Substitute leg, exercising the population the card is actually about

Mutation: invert H9's fireable-Restart-when guard — one token, in a block with
14 direct sites and 6 more behind h9NoLine. Every H9 fixture with no fireable
exit now returns null: "the row goes clean", the exact regression class the
card describes. Expected direction fixed before running: before → TypeError
abort; after → named failing cases, suite completes.

BEFORE (pristine base 7986d973f), verbatim:

file:///…/scripts/pm/check-half-states.mjs:8551
t('H9: …and the finding prescribes the close default', h9OnHoldNoRestartWhen(hold('parked')).includes('not planned'), true);
^
TypeError: Cannot read properties of null (reading 'includes')
at selfTest (file:///…/scripts/pm/check-half-states.mjs:8551:95)

→ cases printed: 0 · named failing cases: 0 · summary line: absent.
The suite reports nothing about the other 1379 cases.

AFTER (this branch):

 ✗ H9: hold with no Restart-when line -> finding (got "object", want "string")
✗ H9: …and the finding prescribes the close default (got false, want true)
✗ H9: closed-upstream form -> clean (got "`pm:on-hold` with its only `Restart-when:` …", want null)
✗ check-half-states self-test: 45 of 1380 case(s) failed.

→ cases printed: 1381 · named failing cases: 45, every one an H9 row ·
TypeError lines: 0. The suite runs to completion and names what broke.

Both legs proved the mutation on disk before running (anchor present → 1,
mutant present → 1, original absent → 0) and restored under
trap … EXIT INT TERM; restore was verified afterwards (git status --porcelain empty, anchor back to 1, mutant 0).

Acceptance (c) — companion sweep of sibling `scripts/pm/*.mjs`

Same two-population audit run over all 13 sibling scripts. check-half-states.mjs
is the only file that carries the shape.

file	nullable fns	pop A	pop B	total
`check-half-states.mjs` (this card)	36	0 (was 145)	0 (was 108)	0
`dispatch-gates.mjs`	8	0	0	0
`check-governed-merges.mjs`	4	0	0	0
`git-history.mjs`	3	0	0	0
`ci-failure.mjs`	2	0	0	0
`check-skill-line-ratchet.mjs`	1	0	0	0
`release-rehearsal-clone.mjs`	1	0	0	0
`bare-root-worklist.mjs` · `check-governed-prose.mjs` · `check-label-desc-cap.mjs` · `check-skill-id-lint.mjs` · `check-governed-queue-guard.mjs`	0	0	0	0
`check-dispatch-gates.mjs`	—	—	—	no `selfTest` function (it drives the tool's own `--self-test`)

The instrument was validated against a known-positive control before the
zeros were believed: run against the pristine base of check-half-states.mjs it
returns popA=111, reproducing the card's headline number exactly. Nothing to
fold, and no finding to file — the sweep's answer is genuinely "only here".

Card 2 — literal stray spaces

Re-derived on this tree with the card's predicate (literal Han + ASCII ,;
: + space + Han, within one line), using the exclusions #12263 used —
fenced regions, YAML front matter, and multi-line 「…」 rulings — taken from the
ratchet's own initialState/advanceState rather than reimplemented.

Result: 7 + 6 + 1 = 14, reproducing the card's table exactly. All 14
deleted; byte delta exactly −14; line counts unchanged, so no line merges,
no CEILINGS row moves, and no table row is touched.

⚠️ Divergence from the dispatch estimate: 14 of 14 fixed, not ~11

The direction anticipated that ~3 of the 14 would be enumeration-marker sites to
leave. Measured, none of them is: the predicate requires a Han
successor, and a circled mark (U+2462 ③ / U+2463 ④) is not Han, so every
enumeration site falls outside the population by construction — as the card
itself noted about its line-end sibling. The judged-deliberate family is
therefore preserved without any of it needing to be excluded by hand.

What makes each of the 14 mechanical rather than stylistic: every one has a
same-line sibling using the identical mark with no space — 实测存在,churn,
判据的,PM, 哈希:空, 它不是,它. The house style is settled on the same line
as the defect.

Per-site judgment — fixed

site	mark	verdict	why
`platform-readings.md:4`	`;`	fixed	mid-sentence prose, no marker follows
`platform-readings.md:55`	`;`	fixed	same line: `存在,churn` — no space
`platform-readings.md:198`	`,`	fixed	mid-sentence prose
`platform-readings.md:274`	`,`	fixed	mid-sentence prose
`platform-readings.md:276`	`,`	fixed	same line: `判据的,PM` — no space
`platform-readings.md:306`	`,`	fixed	mid-sentence prose
`platform-readings.md:313`	`;`	fixed	same line: `盲区:窗口` — no space
`os-dev.md:210`	`,`	fixed	mid-sentence prose
`os-dev.md:232`	`,`	fixed	same line: `推进它:` — no space
`os-dev.md:253`	`,`	fixed	same line: `哈希:空`, `失败;不匹配` — no space
`os-dev.md:263`	`,`	fixed	same line: `三种处置,不是` — no space
`os-dev.md:337`	`:`	fixed	mid-sentence prose
`os-dev.md:376`	`:`	fixed	same line: `它不是,它` — no space
`SKILL.md:127`	`:`	fixed	mid-sentence prose

Per-site judgment — kept

site	shape	verdict	why
`platform-readings.md:56`	`;` + U+2463	kept	enumeration marker — judged deliberate house style; outside the predicate (successor is not Han)
`platform-readings.md:275`	`;` + U+2462	kept	enumeration marker — same
`SKILL.md:883`	`;` + U+2462	kept	enumeration marker — same
`platform-readings.md:26`	`;` + inline code	kept	successor is inline code, not Han; a space between CJK prose and a code span is intended spacing
`os-dev.md:230`	`:` + inline code	kept	same
`os-dev.md:236`	`,` + inline code	kept	same

⛔ No new gate, per the direction: hanAsciiPunctTail (landed in #12263)
already prevents new line-end sources, literal authorship is rare, and the
residual risk is accepted and recorded on the card rather than priced into a
permanent scanner.

Gates

Derived from the real change set — node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack, which read the 4 changed paths from the merge base
itself (three-dot) rather than from a hand-written list — and run at the final
commit c5e68a4d5. Exit codes captured by redirecting to a file first, never
through a pipe. 24 families matched; all 24 green, each quoted from the
gate's own verdict line:

gate	verdict line
`check:pm-half-states`	`✓ check-half-states self-test: 1380 cases pass.`
`check:pm-skill-ratchet`	`✓ …os-dev.md is 461 lines (ceiling 470; headroom 9).`
`check:pm-skill-ratchet`	`✓ …platform-readings.md is 314 lines (ceiling 314; headroom 0).`
`check:pm-skill-ratchet`	`✓ …pm-dispatch/SKILL.md is 1005 lines (ceiling 1005; headroom 0).`
`check:pm-skill-ratchet`	`✓ …pm-dispatch/SKILL.md: widest table row is 765 bytes (pin 765; headroom 0).`
`check:pm-skill-ratchet`	`✓ …os-dev.md: widest table row is 0 bytes (pin 0; headroom 0).`
`check:pm-skill-ratchet`	`✓ …platform-readings.md: widest table row is 0 bytes (pin 0; headroom 0).`
`check:doc-formula-expressions`	`✓ …22 record-scoped formula example(s) across 422 files / 1449 TS blocks judged clean`
`check:agent-model-declared` · `check:agent-test-spelling` · `check:bash32-floor` · `check:cli-command-ids` · `check:cross-package-test-inputs` · `check:doc-authoring` · `check:entry-guard` · `check:nul-bytes` · `check:parse-guard` · `check:partof-closing-keyword` · `check:pm-governed-merges` · `check:pm-governed-prose` · `check:pm-skill-id-lint` · `check:pnpm-filter-targets` · `check:skill-frame-sync` · `check-ci-filter-parity` · `check-closing-keyword-parity` · `check-cross-package-test-inputs` · `check-governed-queue-guard`	exit 0

The MAX_TABLE_ROW_BYTES pins merged in #12263 are respected: the two files with
a 0 pin have no table rows, and SKILL.md sits at 765/765 — unchanged, since
this diff deletes one space from a prose line and touches no table row.

Two gates initially reported red and both said so themselves: the
doc-formula-expressions gate exited on an unbuilt-workspace prerequisite
("Nothing was measured … It is NOT a finding"), green after building
@objectstack/formula and @objectstack/lint; and check-partof-closing-keyword.mjs
refused for want of PR_BODY ("a wiring or usage failure, NOT a verdict") — it
is CI-wired to this PR's body and was additionally run locally against this text
before publication.

skip-changeset: the diff is .claude/** plus one scripts/pm/ self-test
harness. Nothing published to npm changes, so no changeset is owed.

Generated by Claude Code

…ssertion through a row-text wrapper The self-test's message assertions read `predicate(...).includes(needle)` on predicates that are three-valued by design (null when clean, a string when the row fires). The moment a change makes such a predicate go clean, that line throws `TypeError: Cannot read properties of null (reading 'includes')` while evaluating t()'s arguments — before t() runs — so the suite ABORTS there, every later case never runs, and the output names a TypeError instead of a row. Generalizes the per-predicate wrapper PR #12272 established for H34 (`const h34row = (...args) => String(fn(...args) ?? '')`) to the remaining 18 predicates, plus the two local nullable helpers (`halves`, `dead27`) that carry the same shape. 145 call sites switched; predicates themselves are unchanged and their `typeof`/`=== null` assertions deliberately stay un-wrapped. Self-test output is byte-identical to before: 1380 cases pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RMTpSRF5CjMmQBFfPtPCwJ

…nullable-.includes population The card's table counted only DIRECT call expressions (`pred(...).includes(`). A second population reaches the same throw through a local const — `const h9NoLine = h9OnHoldNoRestartWhen(...)` then `h9NoLine.includes(...)` — which a call-expression scan cannot see. Measured on the pristine base: 108 such sites across 29 variables, spanning four predicates the card never named (h20, h21, h23, h28). This half is what actually kept the suite aborting: with population A alone fixed, the H9 ablation still died at the first `h9NoLine.includes`. Two shapes, chosen per variable: - no nullability assertion on the const -> wrap the DECLARATION in place (19 variables, zero call-site churn) - `typeof X` / `X === null` asserted -> keep the raw const and add a `XRow` sibling (10 variables, 56 sites switched), because stringifying in place would make those assertions trivially true Audit now reports 0 unsafe receivers in both populations. Self-test output remains byte-identical: 1380 cases pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RMTpSRF5CjMmQBFfPtPCwJ

…jacent ASCII mark #12081 measured a LINE-END event (a soft break after an ASCII , ; : between Han characters). A faithful re-wrap preserves that space as a literal ", " in source, where the line-end predicate can no longer see it — so the defect migrates rather than disappears. Re-derived on this tree with the card's predicate (literal Han + ASCII ,;: + space + Han, one line, excluding fenced / front-matter / multi-line ruling regions via the ratchet's own state machine): 7 + 6 + 1 = 14, reproducing the card's table exactly. All 14 deleted. Each site has a same-line sibling using the identical mark with NO space (实测存在,churn · 判据的,PM · 哈希:空 · 它不是,它), which is what makes the judgment mechanical rather than stylistic. The enumeration-marker family (semicolon + space + a circled mark such as U+2462 / U+2463) is left untouched — measured, it falls OUTSIDE the predicate by construction (the successor is not Han), so all 6 such sites survive unchanged. Byte delta is exactly -14; line counts are unchanged, so no CEILINGS row moves and no table row widens. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RMTpSRF5CjMmQBFfPtPCwJ

claude added 3 commits August 25, 2026 23:43

claudeBot added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Aug 26, 2026

claudeBot requested a review from os-zhuang August 26, 2026 00:02

github-actionsBot added size/l documentation Improvements or additions to documentation labels Aug 26, 2026

os-zhuang mentioned this pull request Aug 26, 2026
[finding] check-half-states self-test: 111 bare .includes() on nullable predicate returns turn a mutation into a TypeError that ABORTS the suite instead of a named failing case #12273
Closed

os-zhuang marked this pull request as ready for review August 26, 2026 00:06

os-zhuang enabled auto-merge August 26, 2026 00:06

os-zhuang added this pull request to the merge queue Aug 26, 2026

Merged via the queue into main with commit 078bdcdAug 26, 2026
35 checks passed

os-zhuang deleted the claude/issue-12273-selftest-nullable-rows branch August 26, 2026 00:34

os-zhuang mentioned this pull request Aug 26, 2026
[PM seat] skills — 🟢 os-steve (session_01JANH3y7qe3MD8aLaLXci8N) #7623
Open

claudeBot mentioned this pull request Aug 26, 2026
feat(pm): H35 — flag a gate-label removal with no dual-carrier clear behind it #12407
Merged

Navigation Menu

Uh oh!

check-half-states self-test: route every nullable predicate row through a wrapper, and delete the 14 literal stray spaces - #12390

check-half-states self-test: route every nullable predicate row through a wrapper, and delete the 14 literal stray spaces#12390
os-zhuang merged 3 commits into
mainfrom
claude/issue-12273-selftest-nullable-rows

claudeBot commented Aug 26, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

claudeBot commented Aug 26, 2026

Card 1 — self-test nullable-row robustness

What the defect is

⚠️ The population is 253, not 111 — the card's table measured one of two shapes

The fix

Acceptance (a) — full self-test green, case count unchanged

Acceptance (b) — the ablation

⚠️ The card's own leg A no longer reproduces, and that is a real result

Substitute leg, exercising the population the card is actually about

Acceptance (c) — companion sweep of sibling scripts/pm/*.mjs

Card 2 — literal stray spaces

⚠️ Divergence from the dispatch estimate: 14 of 14 fixed, not ~11

Per-site judgment — fixed

Per-site judgment — kept

Gates

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Acceptance (c) — companion sweep of sibling `scripts/pm/*.mjs`