fix(attribution): the shapes a footer collapses to, and the half no hook reaches

[protonspy]

What changed

A session told not to sign its work does not stop signing it — it rephrases. Three escapes were measured walking past the attribution validator, and one surface was never read at all.

A vendor product link is a signature wherever it sits. The old gate asked whether the URL stood alone on its line; a harness that kept the link and moved it into a sentence was read as prose. A product page is nobody's citation, so its position says nothing. Documentation keeps the old gate — an argument that cannot cite the vendor's own docs is one nobody can check, so a docs host or a docs path is reported only standing alone.

Crediting a tool in one's own words is the same claim as the footer. "thanks to", "with the help of", "powered by", "pair-programmed with", and the instrumental voice — "used Claude Code to draft the tests". The last needs a stricter test, since the passive reading is ordinary prose about a harness this repo supports: the name has to be the verb's own object, and not possessive.

scc validate --pr now reads the comments and the review bodies. A commit message passes through commit-msg on the machine that wrote it; a comment is posted straight into the forge, by a tool as readily as by a person, and nothing runs when it is. The comment query degrades on its own, so a forge that answers about the pull request and not about its comments still gets the title and body checked.

What the review found

Two defects in code this change did not write, but did newly expose to live traffic.

signed accepted any Unicode symbol, and > and | are Sm. Every blockquoted line and every table row opened a signature — so the rule reported > Claude Code, Codex, and opencode, which is the exact sentence its own paragraph promises never to report. A review bot writes entirely in blockquotes, and reading comments is what made that live. The symbol now has to be non-ASCII, which is what a signature's decoration actually is; ASCII's symbol block is Markdown punctuation.

Measured over this repository's own record — 147 files, every PR body, issue comment and review comment:

	hits	false
before	56	2 — a bot's quoted file list on #6, and a table row in #6's body
after	54	0

A file is not a name. CLAUDE.md is a path the vocabulary matches on, so a line naming it read as somebody signing the work.

clip escapes now. It formats the offending line into a finding that finding.Set.Report prints to a real terminal on every push, and under --pr that line is written by anybody who can comment — not only by whoever opened the PR. A raw ESC surviving the trip is a control sequence the reader's emulator obeys, enough to scroll a finding off the screen and paint a clean report over it, which is the one lie this validator exists to prevent. Non-graphic runes become spaces, and the cut is by rune so hostile text cannot leave invalid UTF-8 in the report.

Review

Gate	Verdict
scc-code-review	blocked → fixed. One blocker (the signed/mention false positive above) and one minor (the rules line had dropped the clause saying --pr, not commit-msg, is what reads the forge). Both addressed. Its "ticked boxes are true" gate was not-run: this work has no plan, so there was no task list to check it against.
scc-security-review	changes-requested → fixed. One high: terminal-escape injection from a PR comment through clip into render. Fixed at clip. It also flagged the same sink as pre-existing for every other validator — recorded as n-0013 rather than fixed here, since nothing else currently embeds untrusted text.

Both were re-run against the gates, not re-reviewed.

Verification

go build ./... · golangci-lint run · golangci-lint fmt --diff · scc validate --checks --pr — 12 validators, 0 findings; 1001 tests at 86.2%, above the 86% floor. pre-commit, commit-msg and pre-push all passed on this branch.

The false-positive claim is measured, not asserted: the corpus table above comes from scanning this repository's own pull request history with the code as it stands here.

Notes

assets.Version 33 → 35. Two template edits landed on this branch, and scc update writes the manifest only when it applies a file — so a managed file already byte-identical to the next template kept a stale hash, and at the version after that would have compared as edited and never been updated again. Recorded as n-0014.

And then it fired on this pull request

Running scc validate --pr against #52 once it was open reported two findings — one on this branch's own commit message and one on this body. Both are quotations: the commit explains the widened rule by quoting the phrasings it catches, and the paragraph above lists them.

That is the failure the package comment names at the top of its own file — "a pattern that could match mid-sentence would fire on a commit message that merely quotes one of these lines, which is exactly what a fix for this finding looks like". Every other rule is anchored against it: a trailer at the start of a line, a badge, a link or a mention alone on one. The footer family had nothing, and the instrumental voice added here had nothing either.

Second commit: an occurrence behind an odd number of straight quotes or backticks, or inside a typographic quotation still open, is skipped — per occurrence, so an unquoted credit beside a quotation is still caught. The apostrophe is not counted, being a possessive far more often than an opener.

What it gives up is a footer written inside quotation marks, which nothing writes and a reader would take as a mention anyway. Re-measured over 149 files — the corpus above plus this branch's commit and this body: 54 hits, all real, nothing else. scc validate --pr is now exit 0 against the live pull request.