agent-monitoring-retro

What problem does it solve? Accumulated agent run and event logs are hard to act on without aggregation; this Skill turns raw monitoring records into a structured weekly retro report so failures, slow phases, and prompt problems actually get identified and fixed. ## Core Features & Use Cases - Automated Report Generation: Runs make agent-monitoring-retro to produce RETRO-<week>.md from runs.jsonl and events.jsonl, updating the retro index automatically. - Guided Retrospective Analysis: Walks through run summary, gate failure breakdown, tier distribution, agent status, summary quality, and slow runs, then fills the Notes section with concrete findings and one action per issue. - Data Integrity Validation: Cross-checks with make agent-monitoring-validate so DONE log entries, run records, and events stay consistent before trusting the numbers. - Use Case: At the start of a sprint, run the retro to discover that a specific gate blocks most runs and that hotfix tickets are not taking the fast path, then commit the annotated report as institutional memory. ## Quick Start Ask the assistant to generate this week's agent monitoring retro report and summarize the findings that need action.