[finding] A private comment-stripper lives in `scripts/`, where neither `check:comment-mask-adoption` nor `check:parse-guard` can see it

[os-elon]

Found while writing scripts/check-widget-option-census.mjs for #12926, whose whole difficulty is separating code from prose. Recording the coverage gap; ⛔ not fixed there, and nothing about the card depends on it.

The gap

Two gates watch this subject, and their populations do not overlap where the finding sits:

gate	population	subject
check-comment-mask-adoption.mjs	ROOT_DIR_WATCH_HINTS = ['packages/**', 'examples/**']	a NEW private comment-stripper
check-parse-guard.mjs	scripts/**	the three TypeScript PARSER entry points

So a private comment-stripper written under scripts/** is inside check-parse-guard's population but outside its subject, and inside check-comment-mask-adoption's subject but outside its population. Nothing reds.

That is the complement of #12307, which was filed for the strippers OUTSIDE scripts/ and produced the adoption gate with the population above. Its ledger's 14 rows are all under packages/**, which is a true statement about the ledger and not about the tree.

The measurement

On 19763458ff:

$ node scripts/check-comment-mask-adoption.mjs --list
# 14 recorded private comment-strippers (shrink-only)      <- all packages/**

# every scripts/*.mjs that rolls its own comment scan and does NOT
# import the shared mask:
scripts/check-stack-collection-maps.mjs

scripts/check-stack-collection-maps.mjs exports maskLiterals(source) -- a hand-rolled left-to-right scanner covering line comments, block comments, the three string forms and regex literals (it carries its own REGEX_PRECEDER), length-preserving. Its docblock records the incident that produced it: a first draft that stripped rather than masked let one unbalanced paren inside .describe('Screen Flows (ADR-0019)') close the schema literal 14 collections early, and the gate then reported 114 deviations, every one of them its own.

⚠️ Stated precisely, because the two are different claims: this is an ABSENCE OF A WATCHER, not a measured disagreement. maskLiterals was NOT diffed against js-comment-mask's scanner here, and by shape it is in the careful family (scanner-decl with regex tracking) rather than the naive-regex one. What is missing is the thing the adoption gate exists to provide -- somebody noticing the next one.

Why it is worth a row rather than a shrug

The adoption gate's own header states the mechanism: the shared module was landed, some callers were converted by hand, and the population that never moved was found again by hand -- three separate cards for one conversion, and this gate is the part that stops a fourth. A stripper the gate cannot see is exactly a caller that will be found by hand later, and scripts/ is where source-scanning gates are written, so it is the likeliest place for the next one.

Options, none chosen here

• Widen the adoption gate's population to scripts/** and ledger maskLiterals with a measurement, as the second round did for the other 19 rows. Cheapest, and it puts the row where every other row is.
• Convert check-stack-collection-maps.mjs onto the shared mask and take the population widening as a follow-up. A conversion is a MEASUREMENT (two of the first three converted were deleting live code, Five comment-mask ledger rows are measured DELETING live code — convert them off their private strippers #12834), so this is not a mechanical edit.
• Declare scripts/** deliberately out of scope in the adoption gate's header. Explicit is better than the current silence, and it is the option that decays.

[os-elon]

🔒 CLAIMED — domain:devx PM seat (#6023), session session_01CPrUz21stTFhJRUirdc4yw, R28. Branch claude/issue-13143-mask-adoption-scripts-population. Assignee written as os-elon; ⛔ the session ID, not the assignee field, answers whose claim this is.

条款② 声明：NO — this is a repo gate's watched population and, at most, a conversion inside one gate script. No published surface, no contract accept/reject behaviour.

⚠️ A correction this seat owes before anything else: its own hot-file notes had this card down as blocked by #13151 on check-doc-authoring.mjs. That was wrong — this card lands on check-comment-mask-adoption.mjs and check-stack-collection-maps.mjs, and both are free. ⇒ ⛔ do not inherit a blocker from a PM note; the file the card names is the file.

⛔ Zone 1 — rulings, ⛔ not re-adjudicable

1. ⛔ Option 3 is refused. "Declare scripts/** deliberately out of scope in the adoption gate's header" is, in the card's own words, the option that decays — it converts today's silence into a sentence that stops being true the first time someone writes a stripper under scripts/. This card exists because a population that never moved was found by hand three separate times; a header note is the fourth hand-find waiting to happen.
2. ⛔ A conversion is a MEASUREMENT, not a mechanical edit. If you take option 2, note the recorded precedent: two of the first three conversions were deleting live code (Five comment-mask ledger rows are measured DELETING live code — convert them off their private strippers #12834). ⇒ a conversion needs its own before/after reading, ⛔ not a diff review.
3. ⛔ Do not report a measured disagreement you have not measured. The card is careful and you must stay as careful: this is an absence of a watcher, ⛔ not a demonstrated defect in maskLiterals. It was never diffed against js-comment-mask's scanner, and by shape it sits in the careful family (scanner-decl with regex tracking), not the naive-regex one.

Zone 2 — PM mechanism assumptions. Measure them; falsifying one is a result.

• A2.1 I assume the two populations still miss each other exactly as described — check-comment-mask-adoption.mjs's ROOT_DIR_WATCH_HINTS = ['packages/**', 'examples/**'] versus check-parse-guard.mjs's scripts/**. Re-derive both from the source, and confirm the gap is structural rather than an artefact of one hint's spelling.
• A2.2 I assume scripts/check-stack-collection-maps.mjs is still the only scripts/*.mjs rolling its own comment scan without importing the shared mask. ⚠️ Re-sweep — the card measured this at 19763458ff and this repo merges ~18 times a working day. If there is a second one now, that is the strongest possible argument for option 1 and you should say so loudly.
• A2.3 I assume the adoption ledger's 14 rows are all under packages/**. The card's phrasing is worth keeping: that is "a true statement about the ledger and not about the tree" — ⚠️ confirm it, because if a scripts/ row already exists the gap is narrower than filed.
• A2.4 If you widen the population (option 1), I assume the widening admits exactly maskLiterals and nothing else. Measure it: report what the widened population newly reaches, and whether any of it needs a ledger row. ⛔ A shrink-only ledger that suddenly gains unmeasured rows is the [finding] check-dual-build-cjs-loads.mjs's recorded MEASURED counts are 2 entries / 2 packages ahead of the tree they were measured on — and because they are floors, nothing will ever go red about it #13128 class arriving by the front door.

⭐ What makes this card worth doing rather than shrugging at

maskLiterals' own docblock records the incident that produced it: a first draft that stripped rather than masked let one unbalanced paren inside .describe('Screen Flows (ADR-0019)') close the schema literal 14 collections early, and the gate then reported 114 deviations, every one of them its own. ⇒ the careful version exists because the careless one manufactured a hundred false findings — and scripts/ is where source-scanning gates get written, so it is the likeliest place for the next stripper.

Zone 3 — a route, ⛔ not a prescription. Measurement wins.

Option 1 (widen the adoption gate's population to scripts/** and ledger maskLiterals with a measurement) looks cheapest and puts the row where every other row is. ⚠️ But A2.4 is the gate on it: price the widening before taking it. If the price is a pile of unmeasured rows, option 2 for this one file plus a follow-up card may be the honest answer — and saying so with the numbers is a complete deliverable.

---

Generated by Claude Code