| name | asking-diligence |
| description | Spawn a read-only conversational analyst over a code-diligence target's warehouse and repo clones for open-ended drill-down into a finding (e.g., "why is this repo a hotspot?", "who carries bus-factor risk for X?"). Use when the user invokes /code-diligence:ask. Distinct from investigating which tool fed which warehouse column — use interpreting-tool-output for that. |
| argument-hint | <target> <question> |
| allowed-tools | Task, Bash, Read |
The user has invoked /code-diligence:ask <target> <question>.
- Resolve
<target> to its settings file at ~/.claude/code-diligence-<target>.local.md. If missing, tell the user to follow skills/ingesting-target/SKILL.md first and stop.
- From the settings file, capture
target_id, target_name, storage.warehouse_path, and storage.repo_clone_root. Export DILIGENCE_WAREHOUSE_PATH=<warehouse_path> for the agent's environment so the warehouse MCP picks it up.
- Use the
Task tool to spawn the diligence-asker agent (defined in agents/diligence-asker.md) with:
- Initial prompt: the user's
<question> verbatim
- Context block:
target_id, target_name, repo_clone_root
- Stream the agent's response back to the user. For follow-ups in the same turn, re-use the same agent invocation via
SendMessage (do not respawn). A fresh /code-diligence:ask call always spins up a new agent so the read-only boundary stays intact.
- The asker is read-only by design — it can
query the warehouse and read repo files, but cannot write narratives or modify the warehouse. If the user asks for a re-narration, redirect to skills/refreshing-target/SKILL.md (which calls the narrator with write access).
Agent boundary details and the warehouse MCP query surface: references/asker-boundary.md.
