Problem
From verification of #13 (DA-5, MEDIUM — devil's advocate):
spec line 230 寫「planning + has PR → /idd-implement #N (continue,plan-mode 不會自己開 PR;若 PR 已開可能是先前 round 留下)」,但這個邏輯既不勸 user 處理舊 PR,也不警告 stale state。
Type
enhancement / actionable next
Why P3
- 如果 PR 是先前 round 留下(planning 退回 re-plan),舊 PR 內容 vs 現行 plan 必然 divergent → 重新 implement 後 PR 變 stale
- spec 用「可能是先前 round 留下」一句帶過,但 Step 5 是 normative output,不該用「可能」推 user 自己想
Recommendation
- Step 5
planning + has PR 改建議 gh pr close N (stale) → /idd-implement #N 或標註 zombie PR warning
- 或加 detection:若 PR 對應 commit 在本地
git log 中無對應,算 stale → flag
Source
surfaced during /idd-verify #13 (DA-5 finding, MEDIUM severity)
Related: #13
Problem
Type
enhancement / actionable next
Why P3
Recommendation
planning+ has PR 改建議gh pr close N (stale) → /idd-implement #N或標註 zombie PR warninggit log中無對應,算 stale → flagSource
surfaced during /idd-verify #13 (DA-5 finding, MEDIUM severity)
Related: #13