What problem does it solve? AI sessions have no memory of prior debugging attempts, so the same failed fixes get proposed again and again. This Skill maintains a persistent, per-bug log file so every hypothesis, attempt, and result is recorded and checked before any new fix is proposed. ## Core Features & Use Cases - Anti-Loop Enforcement: Requires reading the existing bug log before proposing any fix; approaches marked as failed are off-limits without explicit user override. - Structured Bug Records: Provides a template capturing symptom, repro steps, environment, hypothesis log, per-attempt entries, and resolution details, plus a central INDEX.md tracking open and resolved bugs. - Cross-Repo Bug IDs: Uses a sequential ID convention (e.g., HG-001, XR-002) with repo prefixes for HealthGuard, health_system, IoT simulator, and cross-repo bugs. - Use Case: A token refresh bug in the admin web app resurfaces in a new session. The AI reads HG-001.md, sees two failed attempts with documented reasons, and proposes a genuinely different approach instead of looping. ## Quick Start Ask the AI to check the bug log for the current issue in PM_REVIEW/BUGS before proposing any fix, and to record each new attempt with its result.
