| 1 | # PR Review Playbook v1 |
| 2 | |
| 3 | ## Purpose |
| 4 | |
| 5 | Keep review useful and fast: catch regressions early without turning review into paperwork. |
| 6 | |
| 7 | ## Review Order |
| 8 | |
| 9 | 1. Behavior risk first (what can break for users). |
| 10 | 2. Data/safety risk second (loss, corruption, security, irreversible ops). |
| 11 | 3. Test adequacy third (what proves the change is safe). |
| 12 | 4. Readability/maintainability fourth (can next person change this safely). |
| 13 | |
| 14 | ## Minimal Reviewer Contract |
| 15 | |
| 16 | - Start from changed behavior, not style trivia. |
| 17 | - Prefer 3-7 high-impact comments over many low-value notes. |
| 18 | - Mark uncertainty explicitly. |
| 19 | - Suggest concrete fix or validation step for each critical finding. |
| 20 | |
| 21 | ## Minimal Author Contract |
| 22 | |
| 23 | - Explain intent and risk area in PR description. |
| 24 | - Include quick validation steps (build/test/manual checks). |
| 25 | - Keep PR scope focused; split unrelated changes. |
| 26 | |
| 27 | ## Stop Conditions |
| 28 | |
| 29 | - No unresolved high-severity findings. |
| 30 | - Risky changes have a clear verification path. |
| 31 | - Reviewer understands why this change exists now. |
| 32 | |
| 33 | ## Anti-Patterns |
| 34 | |
| 35 | - Review by checklist only, without reading behavior impact. |
| 36 | - Endless style comments while missing regression risk. |
| 37 | - “Looks fine” with no test or verification signal. |
| 38 | |
| 39 | |