Use this checklist after completing Modules 06–07 to confirm you are ready for the Expert level.
- Can state the 3 conditions that make Agent mode the correct choice over Ask or Plan
- Can name all 4 required anatomy fields: purpose, constraints, tool permissions, handoff criteria
- Can pass the completeness test on any persona definition — all 4 "can I tell you" questions have specific answers
- Can name all 10 agent roles and state the primary output for each
- Can fill in the full permissions matrix for Code Reviewer and Planner without referencing the table
- Can write a complete 3-part handoff prompt from Planner to Implementer for a given scenario
- Can name all 3 failure modes: context pollution, over-delegation, unbounded scope
- Knows which roles are evaluation-only (no Edit access) and why
- Has studied the 10 reference definitions in agents/ — can explain why Code Reviewer has ❌ Edit files, why Planner has ❌ Edit and ❌ Terminal, and why Security Reviewer has ❌ Edit but ✅ Web search
- Module 06 summary Self-Check complete
- Can define a multi-agent workflow as an ordered sequence of bounded sessions, each producing a named artifact
- Can state all 3 properties of a well-formed sub-task: bounded scope, single acceptance criterion, no compound steps
- Can select the correct workflow pattern for Feature Delivery, Bug Investigation, and Refactor and Validate
- Can write a complete workflow file with all 5 required sections before running any session
- Can state all 4 workflow failure modes: one-session sprawl, no artifact discipline, skipped review, workflow drift
- Uses the default model for Planner, Implementer, Code Reviewer, and Test Engineer
- Knows that all outputs must be committed files — nothing exists only in chat history
- Has studied the complete Feature Delivery workflow example in Module 07 and can reproduce its 5-section structure
- Module 07 summary Self-Check complete
Before moving to Expert level, confirm these behaviors:
- I write a workflow file before opening any Agent session — not after
- I stop a session at the exit condition, not when the agent runs out of new ideas
- I treat Code Reviewer as a required step, not optional when confident
- I would not accept a sub-task breakdown that contains "and" connecting two implementation steps
- I commit every session's output before closing the session
→ LEARNING_PATH.md — Path 4: Expert → Start with Module 08 — Advanced Features