What it scores

Category	Weight	What it checks
Instruction files	25	CLAUDE.md, AGENTS.md, Copilot instructions — present and useful?
Docs hygiene	20	README accuracy, stale docs, missing runbooks
Context bloat	25	Token waste: lockfiles, build output, vendored code leaking into context
Secrets & PII	15	Credentials sitting in files your agents can read
Skills & MCP	15	Reusable skills, model-context-protocol wiring

What it generates

CLAUDE.md, AGENTS.md and Copilot instructions — one sweep, every assistant's format
.aiignore — keep the bloat out of context windows
Skill stubs and a Redis + vector-DB memory scaffold (docker-compose included)
AI permissions manifest — governance your security team can read

Runs fully local. Your code never leaves your machine.