The Agent Skills Directory

[SAFE]: The skill operates entirely within the user's project environment to generate documentation. It uses standard tools like Grep, Glob, and Bash to inspect the codebase and write markdown artifacts.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it is designed to ingest and summarize untrusted data from a codebase.
Ingestion points: agents/scanner-agent.md and agents/concept-extractor-agent.md read source code and configuration files from the local project.
Boundary markers: The agent prompts do not currently specify the use of delimiters or 'ignore' instructions when reading codebase content.
Capability inventory: The orchestrator can execute Bash commands via the allowed-tools to write documentation files and check git history.
Sanitization: Content read from the codebase is not explicitly sanitized, but the skill implements a multi-agent review cycle with a critic-agent to verify documentation quality and accuracy before final delivery.

docs-writing