What problem does it solve? Over time, a project's design corpus — stories, decisions, concepts, and surface intents — drifts out of sync with the actual codebase, and no one notices until the gap causes real damage. This Skill runs a periodic, evidence-based audit that re-verifies every claim against the current code and released product, so stale documentation is caught and ruled on rather than silently trusted. ## Core Features & Use Cases - Surface intent and extraction: Co-authors the public-surface intent with the owner in a short interactive stage, then dispatches an autonomous extractor subagent that walks the code and deployment configuration to write the run's surface extraction. - Two-track determination: Measures stories and cold-synthesized user assumptions by driving the released product through public surface elements, while reading decisions adversarially and concepts as vocabulary against the code — all as parallel tasks in a shared task run. - Terminal judge and committed record: Routes every escalation to a single judge, files confirmed gaps as intake issues, writes a run report to the archive, and commits and stamps the audit corpora so currency is a git question. - Use Case: A project owner runs /audit quarterly; the run re-verifies all stories through the product's public surface, finds two decisions contradicted by recent refactors, files them as intake issues, and commits a stamped audit corpus the /document ceremony can build on. ## Quick Start Invoke the /audit slash command in a project with the .ok-planner estate present to start the interactive intent stage and run the full audit.
