← ClaudeAtlas

double-checklisted

Four-eyes principle — verify we didn't screw up this session
TomasLicek/ai · ★ 1 · AI & Automation · score 80
Install: claude install-skill TomasLicek/ai
# Double-Check Spawn a subagent to verify the session. Not a summary — a verification. Senior independent checking we didn't go off the rails, did not misjudge, or forget to clean after ourselves. ## Instructions for subagent You're the second pair of eyes. Review the entire session — not just code, but thinking, process, decisions, and artifacts. Your job is to catch what they missed while deep in implementation, conversation and reasoning. **Verify:** 1. **Did we solve the actual problem?** Or did we drift and solve something adjacent? 2. **Reasoning integrity** — did the logic hold up, or did we talk ourselves into something shaky? 3. **Stayed on track?** Identify where we got derailed. Was the detour justified or did we lose the thread? 4. **Unfinished threads** — things said "we'll come back to" and never did 5. **Session hygiene** — temp files, debug logs, console.logs, hardcoded test values, scratch comments left from debugging 6. **Subagent drift** — did spawned agents go off-topic or produce results that were blindly accepted? 7. **Code changes** — the classic four-eyes: bugs, edge cases, error handling, broken ripple effects, things any reviewer would flag 8. **Scope** — did more than asked? Did less than needed? 9. **Omission audit** — what's NOT there that should be? Check against the original ask: - Requirements mentioned but never implemented - Edge cases discussed but not handled - Config/docs that should have been updated but weren't - Validat