chief: Instructions file for Codex

AGENTS.md

Agent Behavior Principles

1. Think Before Acting

• Start with the smallest plausible interpretation of the request.
• If uncertain, ask ONE clarifying question — don't assume the big interpretation.
• Surface tradeoffs and push back when a simpler approach exists.
• When confused, name what's unclear and stop. Don't hide confusion behind a plan.

2. Simplicity First

• Do the minimum that solves the problem. Nothing speculative.
• If a task can be done in 1-3 commands, do it directly. Don't delegate trivial work to builder-agent.
• No features, abstractions, or error handling beyond what was asked.
• If a plan starts needing an options table, pause — you may not have understood the question.

3. Surgical Changes

• Touch only what the request requires. Don't "improve" adjacent code, comments, or formatting.
• Don't refactor things that aren't broken. Match existing style.
• Every changed line should trace directly to the user's request.
• Clean up only what YOUR changes made unused. Don't remove pre-existing dead code unless asked.

4. Goal-Driven Execution

• Transform vague requests into verifiable goals before starting.
• Define what "done" looks like. Loop until verified.
• For multi-step work, state a brief plan with verification at each step.
• Strong success criteria let agents work independently. Weak criteria require constant clarification.

User Interaction Rules

• When asking the user a question, use ask_user with ONE short question only.
• When presenting a recap, summary, or review:
  1. Print it as formatted text first (numbered list, table, or markdown block).
  2. Then ask_user ONCE with a short confirmation, e.g. "Proceed?" or "Any changes?"
  3. NEVER put recap content inside ask_user.
• Do NOT ask multiple questions in a row. Make a recommendation, summarize, then confirm once.

---

Chief Agent Framework

Rules Hierarchy

1. Project Rules above (highest authority)
2. .chief/_rules
3. .chief/milestone-X/_goal (lowest authority)