What problem does it solve? Expert documentation generated by AI can pass format checks while still being incomprehensible to users who lack code context. This Skill provides an independent, user-perspective acceptance review of expert-team assets, catching cases where the author understands the content but readers cannot. ## Core Features & Use Cases - Cold-Read Testing: Reads documentation without source code first, simulating a zero-context user, then verifies against source code to distinguish documentation problems from misreadings. - Dual-Track Review: Combines usability tests (cold-read questions, terminology self-sufficiency, example reproducibility, navigation reachability) with format compliance checks (R1-R7, CR1-CR9, INDEX rules). - Graded Reports with Auto-Fix: Outputs P0/P1/P2 prioritized findings, automatically repairs mechanical format issues, and routes content-level problems to expert-team or expert-lookup. - Use Case: After expert-team finishes generating documentation for a module, trigger this Skill to verify a new team member could understand and use the docs without reading the source code. ## Quick Start Ask the AI to audit the expert assets for a specific module by saying "审查专家资产" or "expert audit" followed by the expert name.
