Repository navigation
docs(rfc): track a PR item until it lands and give the forced floor wake live facts - #36753
Conversation
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Docs-only RFC with a bounded, phased, revertible design; nothing blocks. Notes
[DESIGN-REVIEWED] 0834510 Description read[DESCRIPTION-READ] 449a48ca82c5cd394ff52984626694af607d91042aa4b734cda6f8302a381558 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
Code change approver recommendation: ApproveWhy: the decider estimates a 10% chance that this change introduces a defect, within the 12% limit for approving on its own, and every check passes. Docs-only RFC plus one index row. I checked its claims about current code against repo/ and they hold: An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
|
Review brief. I read the whole diff (the new RFC and its README row). It is docs only, and all 96 checks are green. I am leaving this as a comment, not an approval, because it is a design decision for the maintainer to sign.
I did not check the code references against main at the stated audit sha. |
bolichen97
left a comment
There was a problem hiding this comment.
Approving. This is the same head I read in full earlier, and all 96 checks are green. It is docs only: the new RFC and its README row. It changes no code and adds no security surface.
Condition for the implementation PR (phase 2, the default bar): items that already exist with a pr_checks bar keep their current meaning and are not rewritten. Only items created after the change default to "merged". The opt-out until: "checks" must be documented in the goal-conductor skill, next to the two lines that now call pr_checks the common case. The RFC's section 7 already says phase 2 rewrites only new writes, so this pins that sentence in the code PR and its tests.
Direction signed by maintainer (bolichen97).
e5bd1c0 to
218e043
Compare
bolichen97
left a comment
There was a problem hiding this comment.
Approving again on the new head. The author rebased onto main after my first approval, and I compared the two saved diffs: the RFC file is the same, and only the neighbouring rows of the README table moved. Checks on this head are still starting and none is red.
The condition from my first approval stands for the implementation PR: items that already exist with a pr_checks bar keep their current meaning and are not rewritten, only items created after the change default to "merged", and the opt-out until: "checks" is documented in the goal-conductor skill.
Direction signed by maintainer (bolichen97).
218e043 to
0834510
Compare
Problem / Motivation
Goal: Record the design that keeps a conductor's PR item open until the pull request merges, wakes the conductor when it lands, and gives a forced floor wake live facts to act on.
A worker pushes its PR green and reports done. The conductor closes the item at green and never learns whether the PR merged: the work-ledger probe reads no PR state. A quiet-floor wake re-sends the same patrol text with no facts, so the forced turn has nothing to check or correct with. On 2026-11-13 items for #36529 and #36302 were closed at green, and #36540 and #36507 were found merged only by floor turns.
Why it matters
The repository owner asked for a capability, not more instructions. Without it, every conductor relies on prompt text to remember to re-check merges, and finds landed work late or never.
Not a goal
What changed (motivation → approach → change)
One new RFC,
docs/request-for-change/rfc-conductor-pr-land-tracking.md, with its row indocs/request-for-change/README.md.pr_mergedis the landed bar with an optionalready: "pr_checks": the worker reports done at green, the item passes only on merge.pr_checkswith arepois stored this way by default (opt-outuntil: "checks").pr_merged+ workerdone+pending, derived at read time. Leaves the goal'sconcurrency_cap, keeps the 32-item store bound.gh_pr's transport. Wakes on merged, closed, conflicting or red, once per fact. About 12 queries an hour per conductor; backs off on rate-limit headers.It covers Ultra Mode and Bot conductors (same probe, ledger and evaluator), lists the tests each phase carries, the GitHub cost, known limits, and a phased rollout with A first. It links
rfc-work-ledger-resource-pressure.md,rfc-conductor-drift-check.mdand #36507.Backwards compatibility
Compatible: docs only; no code reads the new document.
Removes nothing: the diff adds one file and one index row.
Tests
Proof: not provable -- docs or config only
./scripts/docs-lint.shpasses locally ("All documentation checks passed").Manual verification
The batched GraphQL shape in §3.4 was run live against #36507 and #36618: it returned state, mergeable, head and
checkRunCountsByState, withX-Ratelimit-Resource: graphqlheaders.Related Issues
Closes #36751
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)