What problem does it solve? Teams running automated pull request pipelines often cannot tell how many changes merged in a single pass versus how many needed costly second passes, or why. This Skill reads the tracker's history of finished pull requests, deterministically classifies each run as clean, hard recovery, loop checkpoints, or unexplained, and ranks the causes of rework by the wall-clock hours they cost beyond a median clean run. ## Core Features & Use Cases - Deterministic run classification: A shell classifier parses agent marker comments, reviews, and timestamps from tracker data to assign each finished pull request a class, with no judgement calls. - Cost-ranked cause analysis: Causes such as base-branch conflicts, unrecorded reviews, and interrupted runs are ranked by excess hours beyond the clean-run baseline, with ties broken by run count. - Structured retro report: Fills fixed report templates covering outcome shares, ranked causes, unexplained second passes, recovery-by-size breakdown, and an optional handoff of the top cause to om-prepare-issue. - Use Case: Ask "why is our pipeline slow" and receive a report showing that 23% of runs were hard recoveries, that base-branch conflicts cost 613 hours beyond clean runs, and that 36 second passes have no recorded cause. ## Quick Start Ask the agent to run a pipeline retro over the last 30 days of pull requests and report what the second passes cost.
