A set of uncomfortable questions to audit and improve your AI agent (Claude Code, Codex, OpenClaw).
Answers will reveal gaps in the agent's knowledge about you and your projects. Or spark useful insights.
Grouped by theme:
- Processes & Automation — what to automate, what friction to remove
- Assumptions & Priorities — where the agent is wrong about you
- Context & Memory — what gets lost between sessions
- Patterns & Projects — connections you don't see
- Feedback & Improvements — rules to write down
- Answer Quality — where responses are too generic
Ask your agent these questions weekly or when feeling inefficient.
Don't demand answers to all at once — pick 3-5 most relevant.
Questions by Chip for the "Human 2.0" workshop
License: MIT