What problem does it solve? Raw telemetry stores (spans.csv, runs.ndjson) are hard to interpret directly. This Skill turns recorded PtP telemetry into a readable report answering "where is time wasted?" — computing headline work-time and wall-time figures, concurrency factors, breakdowns, and top time sinks per epic or story. ## Core Features & Use Cases - Headline figures: Computes aggregate work time, elapsed wall time, and a concurrency factor per reporting scope, never summing across epics. - Breakdowns and top-N lists: Groups time by phase, agent role, span kind, and tool class, and lists the slowest spans, slowest tool classes, and costliest repeated tool calls (N = 10). - Selector-scoped reporting: Accepts epic:all, epic:XXXX, story:NN, or bare change ids via the shared change-selector grammar, with an empty argument defaulting to all active epics. - Bounded retention pruning: Deletes aged raw telemetry files according to the configured retention window while leaving derived stores untouched. - Use Case: After several pipeline runs, ask for a telemetry report on epic 0032 to see which phase and tool class consumed the most time and which repeated Bash calls were costliest. ## Quick Start Ask the AI to run a telemetry report for a specific epic, for example: generate a telemetry report for epic:0032 showing where time was spent.
