investigatelisted

# Investigate Skill Scopes a problem through structured codebase research and forms an honest assessment. **This skill is investigation-only — it never modifies files, runs tests, or implements anything.** The investigation may conclude that no action is needed; task creation is a conditional outcome, not a guaranteed one. ## Step 0: Start Cost Tracking Record the start of this investigation so cost can be captured at the end: ```bash tusk skill-run start investigate ``` This prints `{"run_id": N, "started_at": "..."}`. Capture `run_id` — you will need it in Step 8. > **Early-exit cleanup:** If any check below causes the skill to stop before Step 8 (e.g., the user never provides a problem statement, or the investigation is abandoned before the report), first call `tusk skill-run cancel <run_id>` to close the open row, then stop. Otherwise the row lingers as `(open)` in `tusk skill-run list` forever. ## Step 1: Capture the Problem The user provides a problem statement after `/investigate`. It could be: - A bug report or error message - A performance concern or regression - A design smell or architectural issue - A feature area in need of refactoring - A vague concern ("something feels wrong in auth") If the user didn't provide a description, ask: > What problem should I investigate? Describe the issue, area of concern, or question you want scoped. If the user does not respond, or declines to provide a problem statement, run `tusk skill-run cancel <run_id>` and stop.