← ClaudeAtlas

investigatelisted

Forensic investigation of a bug, incident, or unfamiliar code area. Use when the user says "investigate", "debug", "trace this bug", or "help me understand this area before I work on it".
rterakedis/leanwheel-skills · ★ 2 · AI & Automation · score 65
Install: claude install-skill rterakedis/leanwheel-skills
# Investigate Skill **Goal:** Reconstruct from evidence; produce case file for cold handoff. Never fix until cause confirmed. **Two modes:** Defect (symptom → root cause) | Exploration (understand area, find gotchas). ## Activation 1. Ask: "What are we investigating?" (error, stack, ticket, log, file/module, description) 2. Determine mode from input. 3. Slug: ticket ID or short name (lowercase, hyphens). Check for existing `docs/investigations/{slug}.md`. 4. Initialize case file immediately (persistent state across interruptions). ## Evidence Grading **Confirmed** (directly observed, cite path:line/timestamp/commit) | **Deduced** (from Confirmed, show chain) | **Hypothesized** (plausible, state what confirms/refutes). Rules: anchor before theories; never delete hypotheses (update status + resolution); document evidence gaps; challenge user description (it's a hypothesis). ## Execution **Defect Mode:** 1. Anchor: one Confirmed piece of evidence. If sparse, go to evidence-light. 2. Standard trace: call paths, recent changes, specific file sections (parallel reads where independent). OR evidence-light: map what should happen, list plausible causes, identify confirmation path. 3. Triage hypotheses: Open/Confirmed/Refuted with evidence + resolution. 4. Root cause: state explicitly, explain symptom, note contributing factors. 5. Handoff: remediation (what, where, story vs. quick patch). Don't implement. **Exploration Mode:** 1. Surface scan: entry points, interfaces (no i