session-critiquelisted

ultrathink # Session Critique: Adversarial Review via Agent Team Spawns an advisor-pattern agent team that adversarially reviews all work done in the current session, surfaces every finding for user approval, and applies only the fixes the user authorizes. ## When to use - Session produced multi-file changes that should be stress-tested - You want adversarial self-critique + code quality review before committing - Changes span documentation, skills, specs, or cross-cutting concerns - You need independent verification that nothing was missed or mis-stated ## When NOT to use - Single-file trivial edit — run `/validate` directly - Standard code review without fix authority — use `/multi-review` - Only need diff/security/schema checks — use `/validate quick` - Work is not yet complete — finish implementation first ## Decision Authority **The user approves ALL non-trivial decisions. No teammate decides autonomously.** Teammates CAN without asking: read files, run verification commands, spawn subagents, report findings. Teammates MUST escalate (via lead → user): applying any fix, dismissing any finding, resolving disagreements, any file content change. ## Procedure ### Phase 0: Scope Detection ```bash # Find this session's commits git log --oneline -10 # Get all changed files git diff <first-session-commit>~1..HEAD --name-only ``` Split files into two ownership buckets (no overlap): - **Bucket A** (self-critic): modifications of existing files - **Bucket B** (code-rev