instinct-debt

Every knowingly-accepted limitation from a Critical-tier decision is marked with an instinct: comment naming the limitation and what would remove it (references/provenance-out.md, case 5). This collects them into one ledger so a deliberate, scoped-for-now tradeoff can't quietly become a permanent, unexamined one.

Scan

Grep the repo for the marker, anchored to a comment prefix so it matches actual inline comments and not prose that merely mentions the convention (a plain grep -rn "instinct:" was tried while writing this skill and returned two dozen false positives — docs, ADRs, this file itself, and the sync-marker's own regex source in hooks/lib.js — none of them real debt):

grep -rnE '(#|//) ?instinct:' --include=*.py --include=*.js --include=*.ts .

(extend the pattern with other comment prefixes — --, ;, <!-- — for the target stack). This also naturally excludes the sync-marker convention (instinct:shared, instinct:host-specific), which appears inside HTML comments and regex source, not as a #///-prefixed inline marker.

Output

One row per marker, grouped by file:

<file>:<line> — <what was limited>. removes when: <the trigger named in the comment>.

Flag any marker with no named removal trigger — those are the ones most likely to rot into permanent, unexamined debt, exactly the failure mode the convention exists to prevent.

End with <N> markers, <M> with no removal trigger. Nothing found: No instinct: debt. Clean ledger. — an expected, normal result for a codebase that hasn't needed the escape hatch yet, not a sign the convention is broken or unused.

Boundaries

Reads and reports only, changes nothing. If a finding warrants action — a marker whose trigger condition has actually arrived — that's a new request through the Gate, not something this report resolves itself.