Fix worktree-rm: detect squash-merged branches via GitHub PR status

[igerber]

Summary

• Replace git branch -d in worktree-rm step 6 with a 3-step approach that queries GitHub first
• Add git fetch origin (best-effort) to update tracking refs before branch deletion
• Add gh pr view check to detect squash merges, rebase merges, and regular merges via GitHub API
• Force-delete (-D) when GitHub confirms PR was merged; safe-delete (-d) otherwise
• Degrades gracefully when gh is unavailable, offline, or no PR exists for the branch

Methodology references (required if estimator / math changes)

• N/A — no methodology changes (slash command config only)

Validation

• Tests added/updated: No test changes (slash command markdown, not executable code)
• Manual testing: verify with merged-PR worktree, no-PR worktree, and gh unavailable

Security / privacy

• Confirm no secrets/PII in this PR: Yes

---

Generated with Claude Code

[github-actions]

Overall assessment: ⚠️ Needs changes

Executive summary:

• P1: gh pr view "$BRANCH" can interpret numeric branch names as PR numbers, leading to a false MERGED result and forced deletion of an unmerged local branch.
• No estimator/weighting/variance/assumption changes; Methodology Registry cross‑check not applicable.
• Doc‑only change; no tests expected.

Methodology

• No findings (no estimator, weighting, variance/SE, identification assumptions, or default behavior changes).

Code Quality

• Severity: P1
  Impact: gh pr view "$BRANCH" treats numeric branch names as PR numbers, so a local branch named like 1234 could be force‑deleted if PR #1234 is merged, even when the branch itself is not merged.
  Concrete fix: Replace the lookup with a head-branch query that cannot be confused with PR numbers, e.g. gh pr list --head "$BRANCH" --json state --jq '.[0].state' and treat empty output as unknown. Update the deletion logic accordingly.
  Location: .claude/commands/worktree-rm.md:L92-L114

Performance

• No findings.

Maintainability

• No findings.

Tech Debt

• No findings.

Security

• No findings.

Documentation/Tests

• No findings.

[igerber]

/ai-review

[github-actions]

🔁 AI review rerun (requested by @igerber)

Head SHA: 2628482c53fa9856697307dc31657e02e92e2bd0

---

Overall assessment: ⚠️ Needs changes

Executive summary

• No estimator, math, weighting, variance/SE, or identification changes; methodology registry reviewed and not applicable for this doc-only PR.
• P1: Force-deleting branches solely based on PR merged state can drop local-only commits (e.g., branch reused after merge or unpushed work).
• Minor robustness: gh pr list --head uses the first PR only, which can misclassify when multiple PRs share the same head name (conservative but inconsistent).
• No tests updated; acceptable for docs, but the new decision logic is behaviorally significant.

Methodology

• P3 (Info) — No estimator/methodology changes in this PR; registry review not applicable. Impact: N/A. Fix: None.

Code Quality

• P1 — Force-delete on PR_MERGED=true can erase local-only commits if the branch has diverged since the merged PR. Impact: potential data loss for work not pushed or not part of the merged PR. Location: .claude/commands/worktree-rm.md:L99-L118. Fix: Before git branch -D, check whether the branch is ahead of its upstream (or has no upstream). If ahead or upstream missing, fall back to git branch -d or prompt. Example logic: git rev-parse --abbrev-ref "$BRANCH@{u}" + git rev-list --count "$BRANCH@{u}..$BRANCH"; only force-delete when count is 0.
• P3 — Using .[0].state can misclassify when multiple PRs exist for the same head name, yielding PR_MERGED=false/unknown despite a merged PR. Impact: branch may not be deleted when it could be, resulting in extra manual cleanup. Location: .claude/commands/worktree-rm.md:L99-L111. Fix: Prefer “any merged PR” or the most recently updated PR (e.g., include updatedAt in --json and select the latest, or check map(select(.state=="MERGED")) | length).

Performance

• P3 (Info) — No performance-impacting changes. Impact: N/A. Fix: None.

Maintainability

• P3 (Info) — No maintainability concerns identified. Impact: N/A. Fix: None.

Tech Debt

• P3 (Info) — No new tech debt identified. Impact: N/A. Fix: None.

Security

• P3 (Info) — No security issues identified; no secrets/PII introduced. Impact: N/A. Fix: None.

Documentation/Tests

• P3 (Info) — No tests updated (doc-only change). Impact: N/A. Fix: Optional—if there is a harness for slash-command behavior, add a small test for merged PR vs. local-ahead branch handling.

[igerber]

/ai-review

[github-actions]

🔁 AI review rerun (requested by @igerber)

Head SHA: 2d6c556addac1cde485729edbb4d044064d4acb2

---

Overall assessment: ✅ Looks good

Executive summary

• No estimator/math/SE/assumption changes; Methodology Registry not implicated.
• Doc-only update to the worktree removal command; logic description is clear overall.
• One minor documentation accuracy nit about what git fetch improves.

Methodology

• No findings. No estimator or inference changes; Methodology Registry review is not applicable.

Code Quality

• No findings.

Performance

• No findings.

Maintainability

• No findings.

Tech Debt

• No findings.

Security

• No findings.

Documentation/Tests

• [P3] Fetch rationale is slightly inaccurate: git fetch origin does not improve git branch -d merge detection (which checks merge into current HEAD), though it can help upstream/ahead comparisons. Impact: could mislead users about why the fetch is useful. Fix: reword to clarify it updates upstream refs for the ahead/behind check, or note it doesn’t affect git branch -d merge detection. .claude/commands/worktree-rm.md:L86-L93