ops_report¶
Generate an execution summary from the indexed P6 event log covering a specified period.
Entry¶
collect
Final output¶
ops_report_output — period, total_runs, success_rate, failure_breakdown, cost_total_usd (always null in current implementation), top_failing_skills, recommendations, summary_markdown.
How it composes¶
Two phases: collect → summarize. In collect, the aggregate.py preprocessor (unsafe mode, 30 s timeout) runs a recall(sources=["events"]) op; if recall returns nothing, it falls back to direct .reyn/events/*.jsonl scanning. The LLM in collect only validates sanity and passes data through. The summarize phase generates the Markdown narrative and recommendations.
Caveats¶
Requires file.read on .reyn/events/ and .reyn/index/. The recall path is only useful after index_events has been run; without it the skill falls back to raw event scanning. cost_total_usd is always null — cost aggregation is not yet implemented.
Usage¶
reyn run ops_report '{"period": "last-week"}'
reyn run ops_report '{"period": "last-week", "focus": "failures", "skills": ["my_skill"]}'
Optional fields: period_days, focus, format, skills.