The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes system tools including git, grep, wc, and ls-files to map the repository, determine file counts, and identify dependencies between components.
[EXTERNAL_DOWNLOADS]: The skill may use npx to fetch and execute the @mermaid-js/mermaid-cli package from the public NPM registry for the purpose of validating Mermaid diagram syntax.
[SAFE]: The skill implements a strict verification protocol where architectural claims are cross-referenced with source code (file and line number). It explicitly avoids freehand generation of diagrams to prevent hallucinations and syntax errors. Additionally, it ensures that all modifications to the user's repository require explicit approval.
[SAFE]: The skill ingests untrusted codebase data (via git and grep) but mitigates indirect prompt injection risks by employing rigid output templates, mandatory source citations for every section, and a verification step that drops any edges that cannot be corroborated in the source files.

blueprint