explore-and-return

name	explore-and-return
description	Broad codebase exploration with fixed deliverable. Use when user asks to explore thoroughly, map modules, trace data flow, or says Return with file tree / mermaid / pain points.

Explore and return

When to use

• "Explore … thoroughly", "map files", "how does X connect"
• Subagent-style tasks with explicit Return: format
• Agent branch / trinity-native-agents structure questions

Deliverable (fixed format)

1. File tree — one line purpose per file (no dump of file contents)
2. Flow — mermaid or bullet data flow (entry → core → output)
3. Pain points — readability, coupling, excessive comments
4. Answer — direct response to numbered user questions

Procedure

1. Glob + Read entry points; Grep imports and callers.
2. Do not paste large file bodies — summarize responsibilities.
3. Use Task/subagent only when breadth exceeds one directory.
4. Keep paths as {REPO_ROOT} when writing notes for export.

Anti-patterns

• Starting implementation before returning the map (unless user asked to implement)
• Listing files without stating purpose
• Ignoring subagent/agent-branch naming the user specified