What problem does it solve? A verification skill's feature map drifts out of date the moment the application changes, leaving stale documentation and broken harness scripts that mislead future verification runs. ## Core Features & Use Cases - Parallel source audit: Launches one read-only subagent per feature file to explain how each feature works from source and flag likely documentation drift with citations. - Mandatory live pass: Drives every feature in a real running session (server, UI, or CLI) following the verification skill's own launch model, with health checks, evidence preservation, and residue cleanup invariants. - Scoped corrections: Ships at most one draft PR of proven doc, harness, or map fixes, editing only the verification skill's own directory and never product code. - Use Case: After a month of product changes, run the maintenance pass to re-verify every mapped feature live, fix drifted descriptions, and report genuine product regressions without papering over them. ## Quick Start Ask the AI to audit the project's verification skill and feature map, exercising every feature live and shipping one PR of proven corrections if anything drifted.