16d-review-binding: gate the three rules that failed under attention

[HackTuah]

Three rules produced during slice 16 had no gate behind them, and each had already failed once while enforced by attention alone. This makes them mechanical.

What lands

1. .githooks/pre-commit refuses a commit on main. Positive match on the full ref, so a tag named main cannot defeat it. A detached HEAD is deliberately exempt — rebase, bisect and the reviewer's index checkout all run detached. A slip-guard, not an authorization control: --no-verify bypasses it and a clone that never ran mix setup has no hook at all. The ruleset is what prevents a push.

2. .githooks/pre-commit greps tracked .md for every path the staged diff deletes or renames, failing on a hit not marked <!-- historical-path --> on the hit line or the one immediately above. Two lines above does not count.

   Earlier drafts of this gate failed open three separate ways, each caught by review and each fixed with a working negative control: C-quoted paths (non-ASCII, tab, quote, newline) that git grep -F could not match; blobs GNU grep classifies as binary, where it exits 0 and emits nothing on stdout; and an enumeration exiting 128, where the affirmative printed over an input that was never searched. It is now NUL-delimited throughout and both enumerations check git's exit status.

   Scope, stated exactly: a tracked path, deleted or renamed by this commit, cited by exact string in a tracked .md. The three defects that motivated the rule are not in that domain — internal/ has never been tracked, one instance is a JSON key and one a stale line number — and the hook says so.

3. tools/signoff.sh binds the signoff to the tree the reviewers read. Exactly two reviewer tree files per round, both equal to git write-tree; .githooks/pre-commit re-checks at commit time. This exists because in slice 16c three of six staged files changed on disk mid-review: the attestation covers index bytes while the drift test hashes worktree bytes, so a signoff written after such a drift attests a diff that no longer exists.

Also: one definition of the reviewable-diff recipe in tools/gate.sh (copies 3 → 2), with a test holding .githooks/commit-msg's remaining copy byte-identical to it until slice 17 removes it; the two tools/gate.sh comment corrections carried from 16b; and BACKLOG.md §11 recording the advisory-print is_uint duplicate.

Review

Four rounds, two independent reviewers each, read-only on checkouts of the index, plus one owner-bounded record-lane read to close.

Round	Tree	Gate integrity	Record / regression	Blocking
1	ebf90b65	FAIL	PASS-with-residuals	2
2	11d2c441	FAIL	FAIL	4
3	a3b8b2ce	FAIL	PASS-with-residuals	2
4	f159ea53	FAIL	PASS-with-residuals	1
read	3797e134	—	CONFIRMED	0

Seven blocking findings. Seven universal quantifiers. Zero mechanism defects after round 1. Every blocking finding was a claim in a tracked file — every, cannot, all three, would have — and not one was a defect in the code those words described. The implementation was correct at the end of round 1 and did not regress under four rounds of adversarial probing.

The slice closed by stripping the last such claim rather than qualifying it: the coverage test is renamed to what it checks, its nine controls kept as regression locks, with no completeness assertion.

Gates

All nine green: compile, format, secret-scan, test (0 failures), credo held at 76, dialyzer held at 43, baseline unchanged, mutation 14 killed / 0 survivors, attestation — and the commit attests to its own diff. derive_diff_hash and attestation verified byte-identical to main's implementation across seven and five ranges, with non-vacuity asserted.

Residuals

Recorded, not fixed: the gate matches an exact path string, so a parent-directory citation is missed and a longer path sharing the prefix is a false positive (fails closed); the detached-HEAD exemption is a two-command route onto main by design; .githooks/commit-msg keeps its recipe copy until 17, bound by the equality test; is_uint stays duplicated (advisory print only, BACKLOG.md §11); one universal quantifier survives at diff_recipe_test.exs:139, out of the closing scope and assigned to 17; and the new example test enumerates via the index but reads via the worktree — the same class this slice closes, bounded by the index-isolation gate, also for 17.

Full evidence, probe results, negative controls and per-round dispositions in internal/slices/16d-review-binding/FINDINGS.md (gitignored, local).

🤖 Generated with Claude Code

https://claude.ai/code/session_01VNdY6wVS773UJgrHUeBN3Y

[HackTuah]

Closing to clear a blocked rollup, not because the work changed.

The rollup carried two check-runs per context — the pull_request run (SUCCESS) and the force-push run (FAILURE) — so Gate - attestation and Gate - baseline may only decrease each showed a failure and the PR went BLOCKED.

Cause, proved: the force-push orphaned d025d35, and tools/gate.sh:173-176 guards the all-zeros sentinel (before on a new branch) but not a well-formed SHA that is unreachable. Line :178 then fails closed — correctly; a gate that cannot measure must not pass.

The commit is unchanged: 27de8a1, tree 3797e134, trailer e72d17e8, signoff written by tools/signoff.sh over the tree both round-5 reviewers read. Re-pushing as a new branch reproduces run 33700784040's conditions, where all nine gates were green.

The CI blind spot is filed separately against slice 18.