Suggest Agent Rules

This analyzes your Claude Code session history and repo structure to generate actionable agent rules, not generic advice. It extracts patterns from actual conversations, looks at what caused rework, checks your git history and PR comments, then outputs concrete do/don't rules with evidence and verification steps. You get a paste-ready ruleset for your .claude directory. Most useful during onboarding, after repeated review feedback, or before major refactors when you want the agent to learn from past mistakes rather than repeat them. The prompt enforces a strict format with confidence levels and verification commands, caps rules at twelve to avoid bloat, and requires evidence for everything.