What problem does it solve? Debugging and code exploration often rely on guesswork and lose context across sessions. This Skill enforces evidence-first investigation, grading every finding as Confirmed, Deduced, or Hypothesized, and persists all progress in a structured case file that another engineer can pick up cold. ## Core Features & Use Cases - Evidence-graded forensics: Every claim is cited with path:line, log timestamps, or commit hashes, and hypotheses are never deleted—only marked Open, Confirmed, or Refuted. - Persistent case files: Investigations are written to a Markdown case file (from a customizable template) covering evidence inventory, timeline, hypotheses, source code trace, and a confidence-rated conclusion. - Dual-mode operation: Handles both symptom-driven defect chasing (tickets, logs, stack traces) and symptom-free exploration of unfamiliar code areas, with delegation to subagents for large file reads. - Use Case: A user reports a crash with a log file. The Skill anchors on a confirmed error string, maps the evidence perimeter, traces causality through the source, and delivers a root-cause diagnosis with a reproduction plan and next-step routing. ## Quick Start Ask the AI to investigate a bug by providing a ticket ID, log file path, or error message, for example: investigate why the battle recorder test fails using the attached stack trace.
