rp-reviews-into-flow

Summary

The design record for itd-28 as re-scoped on 2026-09-21: every review names the commit it read, and the status board says how stale it has become. No new store, no new scrubber.

Scope

1. The pin: every path in abcd that files a review folder under .abcd/work/reviews/<date>-<scope>/ writes review_of_commit: <full sha> into 00-summary.md's frontmatter, taken from the tree it reviewed; the gate receipts keyed by sha already carry the pin in their directory name (criterion 1).
2. The lint: the reviews-charter rule (RD001's sibling) refuses a review folder written after this ships without the key, and names a folder from before it as legacy rather than refusing (criterion 1).
3. The board: the bare abcd status board gains a reviews block: scope or spec id, review_of_commit short sha, commits since (git rev-list --count <sha>..<default>), and a flag past twenty; ordered stalest first (criteria 2, 5).
4. The name: a review of a spec is filed as <date>-<spc-N>-<slug>/, so the spec id is in the folder name; the board reads it from there (criterion 3).
5. The scrub: unchanged; the scanner's store-before-commit redactor and the pre-commit name-guard already run on the reviews tree (criterion 4).

Out of scope

• A review store of abcd's own, a JSON sidecar, a two-stage redaction, a pre-commit verifier: dropped by decision 1.
• Re-running a stale review: the board flags, a person or a run re-runs.

Approach

internal/core/record already reads the reviews tree for the charter check; the pin is one more frontmatter key it reads and the lint one more rule beside RD001. The board's block is computed in internal/core/positioning (where the status board's other blocks live) from the same read plus one rev-list --count per folder through gitutil. The writers are the few places abcd itself files a review (the semantic-gate receipts already pin; the intent-audit ingest and the reviews-charter template gain the key).

How the criteria are satisfied