What problem does it solve? Teams adopting AI coding agents often cannot tell whether their surrounding workflow mechanisms (rules, skills, verification, review, knowledge capture) actually work or merely exist as files. This Skill audits the AI collaboration workflow of a project and produces an evidence-graded health report instead of gut-feel judgments. ## Core Features & Use Cases - Five-dimension assessment: Evaluates task understanding, controlled execution, change verification, reliable delivery, and learning retention, each with three concrete check items. - Evidence-state ladder: Every conclusion is labeled as exists, wired-in, used, effective, missing, unobserved, or not applicable, and scores are capped by the strongest evidence found. - Prioritized findings report: Outputs a Markdown report with per-dimension scores, check-item details, and severity-tagged findings that each include impact, minimal fix, and verification steps. - Use Case: After inheriting a project with dozens of skills and rules, ask for a workflow health check to discover that a design-review skill exists but has never produced a single review artifact, then get a concrete fix routed to the right mechanism. ## Quick Start Ask the agent to run a harness review on this project and produce a workflow health report with evidence status for each dimension.
