docs(engine): record the file:// root-widening decision — order 20260818T132719Z

Answers the order's two asks, verified independently:

1. "sjekken leser bare README.md" — FALSE against the current codebase.
   inspectRepo has scanned every tracked .md file since 2026-07-27 (commit
   816ba97), three weeks before this order. Reproduced the order's exact
   scenario against a fresh clone of open/app-creator: CLAUDE.md:136 does not
   fire, but not from the diagnosed cause — the cited link uses the `/.../`
   elision this engine already treats as a documented placeholder convention
   (the same "18/22 split" the FILE_URL_LEAK comment already describes), not
   a leak. Injected a real (non-elided) file:// leak into the same CLAUDE.md
   in the same clone: LINK-FILE-URL fired immediately, correct file:line.
   No code change corresponds to this half of the order.

2. "bør regelen gjelde file://-lenker generelt, ikke bare /Users|home" — live,
   unmeasured, answered here. Grepped every tracked .md file in the 21 of 22
   registered repos with a local clone (llm-security-commons absent).
   Positive control: the two real /Users/ leaks in ki-produktivitetsmodell's
   CLAUDE.md were found, proving the query isn't silently empty. Result:
   zero occurrences of any other root (no C:\, /private/, /var/, /tmp/) in
   150 file:// hits across 12 repos — every other hit is a generic
   placeholder already declined by the Users|home anchor. Recorded as a
   RELEASE-ASSETS-shaped rejection in the code comment: no subject, not a
   low rate. Widen the day a real one appears.

Closing the order with --no-commit: this resolves both halves, it does not
hand anything back for the sender to act on.
This commit is contained in:
Kjell Tore Guttormsen 2026-08-18 17:02:27 +02:00
commit 5fe1743105

View file

@ -1368,6 +1368,20 @@ function isFixturePath(path) {
// A home directory is what makes a `file:` URL a leak rather than a scheme the
// gate declines to resolve. Anchored on the two roots a real machine path
// starts with; a bare `file:///abs/path.html` placeholder is not one.
//
// WIDENING TO OTHER ROOTS WAS CONSIDERED AND REJECTED FOR NOW — an order
// (`.claude`, 2026-08-18) asked whether `C:\`, `/private/`, `/var/`, `/tmp/`
// and similar roots deserved the same anchor, having measured only the
// `/Users/ktg` case. Measured here across every `.md` file in the 21 of 22
// registered repos with a local clone (`llm-security-commons` absent):
// exactly two REAL leaks exist in the entire corpus, both `/Users/...` in
// `ki-produktivitetsmodell` (already caught — the positive control that
// proves this grep can find one). Every other `file://` hit is a generic
// placeholder (`/abs/path`, `<abs...>`, `${...}`) that anchoring on Users|home
// already declines to flag. Zero occurrences of any other real root, in this
// corpus, on this date — a RELEASE-ASSETS-shaped rejection: not a low rate to
// veto later, an absent subject. Widen this the day a real one appears, not
// before; re-measure rather than trust this count if this comment goes stale.
const FILE_URL_LEAK = /^file:\/\/\/?(Users|home)\//i;
// Relative file links only. Anchor resolution depends on per-renderer heading