fix(context-recovery): look up <branch>-reviews.jsonl with the canonical slug form, not the raw branch (#1127)

[harjothkhara]

Summary

The Context Recovery preamble block (present in every skill) reports REVIEWS: N entries by probing $_PROJ/${_BRANCH}-reviews.jsonl, where $_BRANCH is the raw git branch --show-current. But gstack-review-log (writer) and gstack-review-read (reader) both name that file with the sanitized slug form from gstack-slug ($BRANCH, restricted to [a-zA-Z0-9._-]).

For any slashed branch — the conventional feat/*, fix/*, chore/* — the two forms disagree (feat/foo → feat/foo vs featfoo), so the current branch's review count was never found. ($_BRANCH isn't even defined inside this self-contained block, making the lookup doubly fragile.)

Closes #1127.

Audit of every *-reviews.jsonl access

Site	Branch form	Role
bin/gstack-review-log	$BRANCH via gstack-slug (strip)	writer ✓
bin/gstack-review-read	$BRANCH via gstack-slug (strip)	reader ✓
ship/SKILL.md.tmpl	$BRANCH via gstack-slug (strip)	writer ✓
Context Recovery block	raw $_BRANCH	existence check ✗
specialist-stats, retro, artifacts-init	* glob	branch-agnostic ✓

So on current main the writers and reader already agree on the gstack-slug form — the Context Recovery existence check was the sole straggler.

Fix

Route the review lookup through $BRANCH (already in scope from the block's eval gstack-slug), matching how reviews are written and read. The timeline.jsonl greps keep using raw $_BRANCH — timeline records the raw branch in its branch field — and the generator now documents the two-form convention inline so they don't get conflated again.

Edited the source-of-truth generator scripts/resolvers/preamble/generate-context-recovery.ts; regenerated all SKILL.md via bun run gen:skill-docs (idempotent — the per-skill diff is exactly the one review line).

Scope

Deliberately converges the one remaining straggler onto the existing canonical sanitizer (gstack-slug). It does not change gstack-slug's output form, so there's no on-disk migration and no conflict with the in-flight gstack-slug PRs. Changing the canonical form itself (strip → dash for readability, as the issue suggests) is an orthogonal, migration-bearing decision left to the maintainer — now a one-line change in a single function.

Tests

test/regression-1127-context-recovery-branch-form.test.ts pins both halves (reviews → $BRANCH, timeline greps → $_BRANCH) for both claude and codex hosts. Green together with preamble-compose, review-log, gen-skill-docs, gen-skill-docs-idempotency, audit-compliance — 415 tests, 0 fail.

🤖 Generated with Claude Code

[trunk-io]

❌ This pull request has been removed from the merge queue because required statuses are not defined in .trunk/trunk.yaml, your repo's branch protection rules for main, or the merge queue specific settings. See more details here.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here

[jbetala7]

Independently reviewed against current main (c43c850c) — looks correct and complete. The split-form handling is exactly right:

• Timeline greps correctly keep $_BRANCH (raw). In the assembled preamble _BRANCH is defined upstream by generate-preamble-bash.ts (_BRANCH=$(git branch --show-current ...)), and that raw value is what the timeline writers record — both the started event in generate-preamble-bash.ts and the completed event in generate-completion-status.ts write the branch field from git branch --show-current (raw). So grepping timeline.jsonl by $_BRANCH matches what was written.
• Reviews lookup correctly switches to $BRANCH (slug). gstack-review-log writes $SLUG/$BRANCH-reviews.jsonl and gstack-review-read reads the same, where $BRANCH comes from gstack-slug (tr -cd 'a-zA-Z0-9._-'). For any slashed branch (fix/foo → fixfoo) the raw and slug forms diverge, so the old ${_BRANCH}-reviews.jsonl probe never matched the on-disk file — confirmed on main.

Collision check: no competing open PR (#1621 also touches the context-recovery block but only its slug-derivation, unrelated to the reviews-filename form). Regression test regression-1127-context-recovery-branch-form.test.ts covers it. LGTM.

[harjothkhara]

/trunk merge

[harjothkhara]

These look ready from my side — the branch is clean/mergeable and an independent review confirmed the fix is correct and complete. The only red check is the Trunk merge-queue gate, which from what I can tell just needs a maintainer to enqueue it (no code/test failure). @garrytan whenever you have a moment, could these be added to the merge queue? Happy to rebase if anything is needed. Thanks!