What problem does it solve? Runtime bugs like crashes, wrong output, race conditions, and intermittent failures are hard to fix because developers often patch symptoms without understanding the root cause. This Skill enforces a disciplined debugging loop so every fix is backed by observed evidence, not guesswork. ## Core Features & Use Cases - Structured Debugging Loop: Reproduce the failure, form at least three ranked falsifiable hypotheses, instrument before modifying, test each hypothesis, fix only after confirmation, and verify adjacent behavior is unchanged. - Concurrency and Environment Diagnostics: Reference guides cover timing-dependent failures, use-after-release patterns, and works-locally-but-not-in-prod environment diffs. - Escalation and Hygiene Rules: After two failed fix attempts, escalate with a full attempt timeline; all temporary instrumentation is tracked and removed before the task is done. - Use Case: A flaky test fails only under load in CI. Use this Skill to form ranked hypotheses about shared-state ordering, force the suspected interleaving with a controlled delay, confirm the cause, and ship a verified fix. ## Quick Start Ask the agent to debug why your application crashes intermittently in production but works locally, and have it follow the reproduce-hypothesize-instrument-fix-verify loop.
