Repository navigation
Session Handoff [auto-2480]: Read Copilot's Collapsible Overview Blocks per #2480 #2547
Description
Next steps, in priority order
- Fix Recognize Copilot's Collapsible "N Open Findings" Overview Block in pr_review.py #2480: teach
scripts/pr_review.pyto read Copilot's collapsible<summary><strong>N open findings</strong></summary>overview block, and, per the fold recorded on Recognize Copilot's Collapsible "N Open Findings" Overview Block in pr_review.py #2480, the "Resolved Since Last Review" block from Recognize Copilot's "Resolved Since Last Review" Overview Block in pr_review.py #2479, through one recognizer with a test for each shape built from constructed text. The pull request body carriesCloses on promotion: #2480, #2479. Done looks likepr_review.py waitreading a review carrying either block without exiting 43.
External blockers
None.
Internal dependencies
Decisions #2533, #2540, and #2546 each came from this block going unread, so this fix is what stops later lanes parking on it. Do not answer those decisions here.
State
No branch or pull request yet. Re-derive from git and GitHub.
The parked decision queue
Not applicable to this lane.
What the last round did
Queued from the unattended run after three rounds parked on this one cause.
What not to repeat
Do not add a per-shape special case beside the existing readers. #2516's rule applies: one recognizer for the block class.
New learnings
None.
Activity
ptr727 commentedon Oct 8, 2026
Scope widened by the maintainer, 2026-10-08, from the
windowslane: #2544 folds into this lane's one-recognizer fix. #2544 is the same class, a Copilotccr-overview-v2section the reader does not parse, where the verdict headline names findings beside**0 open findings**. The recognizer should report that headline as its own reading, not as a clean pass. Add a test for the constructed shape in #2544, and add#2544to the pull request'sCloses on promotionline.ptr727 commentedon Oct 8, 2026
Parked, waiting on decision #2564
Done. #2562, from
feature/auto-2480intodevelop, is open and pushed atdc1de527. Its changes:read_overviewinscripts/pr_review.py, reads the second overview format. It covers the verdict and headline, the count-first open-findings section in both numbers, both resolved spellings, andOpen (N). The rows live in one table,OVERVIEW_SECTIONS.fullDatabaseId.50.scripts/README.md, the module docstring, andpr-review-conductare updated, and the generated skills are rebuilt.The pull request closes on promotion #2480, #2479, #2544, and #1723. #1723 is included, because its three shapes were already vetted and this change gives
Open (N)its reader. #2481 is not included and is left to its own lane. It would change whatoverview=T/Mmeans across the docstring, the README, and the skill, which is past one recognizer.Review state.
dc1de527, recorded 0 findings.review_on_head=local). When this lane parked, checks were still running (6 of 10).pr_review.pyreads every round asshapes=ok. Develop's copy still exits 43 on the(N) open finding(s)summaries, which are the shapes this pull request fixes.Why it parked. Copilot's second round has the headline "Two moderate issues remain in unlinked-entry counting and clean-verdict headline detection." Its one open finding was the first issue, which is now fixed. The second issue names #2544's headline-sentence arm, which #2562 deliberately does not implement. Leaving an acceptance arm out is a value call, so it is decision #2564. That issue sets out the options and a recommendation.
What remains.
pr_review.py comment, citing Ship the Exit-50 Verdict Reading Without #2544's Headline-Sentence Arm? #2564. Then confirm that checks are green and squash-merge todevelop. This run's scope ismain, so thedeveloptomainpromotion follows.Worktree.
/home/pieter/repos/worktrees/ProjectTemplate-auto-2480is clean and is left standing onfeature/auto-2480.Lessons. None filed.
ptr727 commentedon Oct 8, 2026
Continued in #2565, round 2 on track
auto-2480.