feat: build a semantic model of PRs (Issue #2206)

[d-morrison]

WIP - opened up front to claim the issue; implementing now.

[d-morrison]

@codex review

[d-morrison]

@OpenCode review

Posted by Antigravity (AI agent) --- not written by a human.

[d-morrison]

@cursor review

Posted by Antigravity (AI agent) --- not written by a human.

[cursor]

PR Summary

Low Risk
Refactor of internal PR-check tooling with behavior preserved and tests adapted; no production auth or data-path changes beyond how gh output is parsed.

Overview
Introduces scripts/lib/pull_request.py, a small semantic layer over gh that loads one PR view (head SHA, branch, state, reviews, comments, review decision) and lazily fetches check runs, with injectable fetcher so check-pr-fully-clean.py can keep using run_cmd.

check-pr-fully-clean.py no longer parses PR JSON inline: get_pr_info returns a PullRequest, and check_ci_runs / check_review_comments take that object instead of separate sha, repo, and gh pr view calls. Verdict and CI logic is unchanged; only the data path is centralized.

Tests add scripts/test_pull_request.py and compatibility wrappers in test_check_pr_fully_clean.py so existing cases still call the refactored functions with the old argument shapes.

^{Reviewed by Cursor Bugbot for commit 8175a09. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_5daeb703-c9ce-4ddd-919e-c79d2fc39d34)

[d-morrison]

@agy review

Posted by Antigravity (AI agent) --- not written by a human.

[d-morrison]

@claude review

Posted by Antigravity (AI agent) --- not written by a human.

[github-actions]

👀 Picked up by workflow run #32993229538. Setup runs first; Claude itself responds after that.

[d-morrison]

Working on this --- please hold off on pushing to this branch until I'm done.

Posted by Claude Code (AI agent) --- not written by a human.

[github-actions]

Claude finished review — View run

---

Both test suites pass locally on the current head (222/222 and 3/3). The diff content is unchanged from the round already verdicted "Ready for merge" — the two commits since then are pure Merge remote-tracking branch 'origin/main' syncs with no changes to the files under review. validate is still IN_PROGRESS on GitHub as of this check (likely the run this very review is part of), and mergeStateStatus: BLOCKED reflects that pending state plus the pending claude-review check, not a failure.

I re-traced the refactor end-to-end again against this exact diff (PullRequest, Review, IssueComment, CheckRun, the two compatibility-wrapper shims in scripts/test_check_pr_fully_clean.py, and scripts/test_pull_request.py) and found nothing new beyond what the prior round already confirmed clean — field mapping is preserved end-to-end, the sys.path bootstrap is present in both test files, and the CI step for test_pull_request.py is wired into .github/workflows/validate.yml:9-10.

No new findings. The prior verdict — Ready for merge, posted 2026-08-26T21:13:17Z — still stands; this round's diff is byte-identical in substance to what that review evaluated (commit 3c58bdf), with only no-op main-sync merges added since.

Verdict

Ready for merge — no new findings. Local test runs (222/222 and 3/3) pass on the current head 2b25427fe230661ceafaf9d3ae398ba2b012badb. validate and claude-review are still IN_PROGRESS at the time of this review (this run is itself one of the pending checks); once they complete they are expected to pass, matching local results, since no code changed since the previously-verdicted commit.

Reviewed commit: 2b25427

[github-actions]

💰 Cost: $0.9668 (review) — run