investigate

To generate an investigation plan from end-of-chat context and documentation, run /investigate to synthesize chat summaries and relevant docs into a defined focus, hypotheses, and a set of actionable tasks.

You can gather evidence without making edits by dispatching a reviewer sub-agent in read-only research mode, which scopes work to safely collect information and synthesize findings without modifying files.

Automating high-level research and task generation after a chat session is best done by identifying end-of-chat context and relevant docs to define an investigation focus, then dispatching a reviewer sub-agent to produce a plan.

The investigate skill requires no external dependencies or components to run, relying entirely on read-only access to chat context, docs, and the correct dispatch channel for the reviewer sub-agent.

You should use a read-only research mode for software engineering investigations when you need to safely gather evidence and synthesize findings into an actionable plan without risking unintended edits to your codebase.

The reviewer sub-agent synthesizes findings into an actionable plan by operating in research mode to gather evidence via read-only access, then compiling that evidence into a final plan and tasks for user presentation.