/decisions · 2026-07-28 · ops · self-review
This ticket has no scope on the Notion page itself — just a title. It traces to a real finding, not a blank slate: the /improve autonomous run on 2026-07-27 queued it from review-log.md (Review 20 cycle) after noticing the newsletter-frontmatter carve-out list has been extended 9 times in 3 months and still keeps missing new doc classes. But nobody has turned that finding into an actual build spec. Honest state: idea exists, spec doesn't.
From ~/rdco-vault/01-projects/self-review/review-log.md (2026-07-27 entry), the proposed fix as described there:
self-review's rubric and audit-newsletter-outputs.py — currently branch on an ever-growing list of hardcoded doc-class exemptions (deep-research briefs, concept docs, X-long-form, YouTube, etc.), extended piecemeal every time a new class slips through.provenance: field at authoring time (e.g. newsletter vs original-synthesis vs build-artifact). Both graders branch on that single field instead of maintaining separate carve-out lists.provenance: onto the ~3,500 existing vault docs, no list of which producer skills would need to start emitting it, and no estimate of effort — none of that has been worked out yet.Nothing else in the vault mentions "opt-in newsletter contract" or "invert the self-review frontmatter rubric" outside this one log entry — this is the whole trail.
Green-light Ray to turn the review-log sketch above into a real spec (value set for provenance, backfill plan, producer-skill list) and bring back a build proposal.
Scope it + sendNot worth a structural change right now — the carve-out-list approach is annoying but working. One-line reason.
Archive + sendTake a smaller concrete first step instead of the full inversion — e.g. just document the carve-out list once and see if that holds for a few cycles.
Split + send