What problem does it solve? Repository documentation is frequently stale or flatly wrong, and onboarding onto an unfamiliar repo wastes hours rediscovering what is actually deployed, which commands really work, and which guards silently fail. This Skill produces a verified mental model of a repo where every load-bearing claim is checked against code. ## Core Features & Use Cases - Parallel read-only analysis: Fans out codebase-analyst agents partitioned by domain (structure, backend, frontend, data, security, deploy) with strict file:line evidence requirements. - Claim verification and reconciliation: Treats existing READMEs and docs as leads, not sources, and resolves agent conflicts by reading the code directly. - CLAUDE.md generation: Writes dated, verified root and per-subdirectory CLAUDE.md files including a high-value Traps section documenting silent failures. - Use Case: When joining a project whose README describes a chatbot but the code is a commerce platform, run this audit to produce accurate orientation docs with cited evidence and a list of misleading traps. ## Quick Start Ask the assistant to audit this repository in depth and generate verified CLAUDE.md files for the root and each significant subdirectory.
