What problem does it solve? Evaluating a real product, dataset, API, or skill under production-like conditions often produces ad-hoc, unverifiable feedback. This Skill standardizes the audit into a moderated, evidence-based process with owner-in-the-loop checkpoints, so every finding is reproducible, classified, and ready for a GO/NO-GO decision. ## Core Features & Use Cases - Two-persona audit loop: a first-person user driver exercises the real artifact surface while an expert observer scores explicit dimensions with P0–P3 severity. - Evidence and governance discipline: every finding cites captured evidence, external mutations follow Draft-First rules with tagged, logged, reversible actions, and uncertain verdicts are confirmed with the executive at finding-time. - Structured deliverables: generates PROMPT.md, TASKS.md, PROGRESS.jsonl, QUESTIONS.jsonl, an ISSUES.md backlog, persona experience narrative, expert critique, and a FINAL_REPORT.md with GO/NO-GO verdict. - Use Case: Before a beta release, ask for a UX and accessibility audit of your staging app; the Skill instantiates the audit template, runs both personas, captures screenshots and logs, and delivers a prioritized issue backlog with a release-readiness verdict. ## Quick Start Audit the staging build of my web app for UX and accessibility issues and give me a GO/NO-GO release verdict with an issue backlog.
