docs: lead decisions 136 (CI changes) and 137 (doc-only findings don't block)

Jason's 17:16Z rulings, relayed by Mos: the #271 CI steps (Vikunja
#273-#278) and Q143. Decision 137 is the reviewer rule for Vikunja #279;
the queue README's verdict paragraph points to it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
This commit is contained in:
2026-10-14 12:24:14 -05:00
co-authored by Claude Opus 5.5
parent 559ef22875
commit 0b106518ad
2 changed files with 48 additions and 1 deletions
+45
View File
@@ -3041,3 +3041,48 @@ which stay with him. Each item names who decided it and what happened.
the task's latest pushed commit. A task with no push needs none.
- Owner Darkwing, reviewers Dewey and Sage (decision 131). Building
starts when row 91 closes and Sage switches the row.
136. **CI changes from the #271 debate (2026-10-14, Jason's ruling at
17:16Z, relayed by Mos).** Jason: "I agree with the recommended steps
for CI improvement." Sage's sage-0178 adoption stands, item 5
included as deferred. Mos filed the work on Vikunja P19:
- #273: builder rounds may be targeted. They run the changed
packages, the packages that import them, and every
`scripts/test-*.sh` that reads a changed path, listed in build.md.
Sage's landing gate stays full.
- #274: the landing gate drops its pkg-queue step, because
`scripts/test-queue.sh` runs `packages/queue/tests/` itself (9 test
files, 156 tests).
- #275: queue revs are committed at a landing, or when a seat needs
them pushed, not after every move.
- #276: a doc-only round needs no gate. A round is doc-only when every
path it changes is a `.md` file that no package test and no
`scripts/test-*.sh` reads.
- #277: the flaky tests get rows, then the worker trial.
- #278: rows that are ready together land through one batch gate.
Rows that share a path go in separate batches.
137. **Doc-only review findings don't block (2026-10-14, Jason's answer
to Q143, "Q143: yes", relayed by Mos; Vikunja #279).** This applies
to every reviewer: Filbert, Dewey, Rocko, Sage and any later seat.
- A finding is doc-only when every path it names passes decision
136's #276 test: a `.md` file that no package test and no
`scripts/test-*.sh` reads. Check with `grep -rlF <basename>
packages/*/tests scripts/test-*.sh`. Any hit makes the finding
blocking. Tests read the PRD (`docs/prd/mosaic-stack.md`) and
QUEUE.md, so findings on them block. A row's brief is pinned by
blob in `queue.json`, so a wrong brief gets a new revision from
Sage, not a follow-up commit.
- Doc-only findings alone don't stop an approval. The reviewer records
`approve` and lists them in the verdict comment under "Doc-only,
follow-up". A round with any other finding records `changes`, as
before, and lists its doc-only findings too.
- The builder fixes them in a follow-up packet after the landing.
Sage lands it as its own commit with no gate (#276), and tracks it
in the row note until it lands. The reviewer checks the follow-up
only if the fix changes meaning, and says so in the verdict.
- Findings on code, tests, contracts, roles, schemas or any file a
test reads still block. A missing test the brief asks for is a code
finding.
- When unsure, the reviewer calls it blocking and says why. Sage can
move it down.
+3 -1
View File
@@ -138,7 +138,9 @@ abandoned one.
reviewer in the current round, so a Jason-gated row reaches Jason only
after its reviewers approved. `done` also refuses `--evidence`. The queue
doesn't fetch the verdict comment; `record` is cooperative like every
other actor claim.
other actor claim. A round whose only findings are doc-only records
`approve`, and the fixes land as a follow-up commit (lead decision 137,
`docs/plans/2026-09-26_lead-decisions.md`, which defines doc-only).
- **`review verify-commit ID REF`** checks a prospective commit or tree
against the current round's candidate. For a manifest, every listed path
must hash to its digest. For a commit candidate, every path the candidate