acquire-codebase-knowledge

The acquire-codebase-knowledge skill maps, documents, and onboards a developer into an existing codebase by producing seven populated documents under docs/codebase/: STACK, STRUCTURE, ARCHITECTURE, CONVENTIONS, INTEGRATIONS, TESTING, and CONCERNS. It solves the onboarding and knowledge-capture problem where a repository lacks trustworthy, verifiable documentation, and it enforces a strict evidence rule: only document what is verifiable from files or terminal output, never infer or assume, marking unknowns as [TODO] and intent-dependent decisions as [ASK USER].

The workflow runs in four phases. Phase 1 runs a bundled scan script (scripts/scan.py, invoked with python3 and an output path) and reads intent documents such as PRD, README, ROADMAP, and SPEC. Phase 2 investigates each documentation area using inquiry checkpoints and a stack-detection reference for ambiguous stacks. Phase 3 copies the provided Markdown templates into docs/codebase/ and fills them in a defined order. Phase 4 runs a mandatory validation loop. The skill advertises multi-language manifest detection, CI/CD and container/orchestration detection, code metrics, and security/compliance config detection, and it supports a focus-area mode to prioritize specific documents. Every document carries an evidence list of concrete file paths, and the final output surfaces numbered [ASK USER] questions and intent-versus-reality divergences.

It targets engineers, tech leads, and teams onboarding to unfamiliar repositories or generating architecture documentation. It requires Python 3.8+ and git, and the scan script is run from the target project root as a read-only inventory step.