Agent Review

Review an agent's AGENTS.md by running two parallel sub-agent checks via TaskCreate:

1. Structure check — frontmatter schema, body coherence, schedule validity, permission completeness
2. Quality check — does the agent fulfil the user's intent?

Workflow

1. Ask the user which agent to review (or infer from context).
2. Locate the agent:
   • Global: ~/.treasure-work/agents/{name}/AGENTS.md
   • Workspace: {workspace}/agents/{name}/AGENTS.md
   • Use agent_get if the path / scope is unclear.
3. Fetch the agent's frontmatter + body via agent_get { name }.
4. (Optional, for agents that have already run) fetch the most recent run via chat_read to look for execution issues that surface only at run time.
5. Launch both checks in parallel using TaskCreate.
6. Collect with TaskGet and present a unified report.