The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts scripts/render_mermaid.py and scripts/validate_mermaid.py invoke the mmdc (Mermaid CLI) tool using subprocess.run. The implementation uses an argument list rather than a shell string, which mitigates command injection risks.
[EXTERNAL_DOWNLOADS]: The skill can download the Chromium browser binary from well-known sources via the pyppeteer library's chromium_downloader when the --install-chromium flag is used.
[EXTERNAL_DOWNLOADS]: Declares a dependency on the pyppeteer Python package and expects the mmdc command-line utility to be available on the system PATH.
[PROMPT_INJECTION]: The skill processes diagram code provided via files or standard input, which constitutes an indirect prompt injection surface.
Ingestion points: scripts/render_mermaid.py and scripts/validate_mermaid.py read Mermaid content from files or standard input.
Boundary markers: Absent. The Mermaid source is passed to the tool without explicit delimiters or instructions to prevent processing of embedded commands.
Capability inventory: The skill can execute the mmdc subprocess and perform local file system read/write operations.
Sanitization: Absent. Input is processed without validation or filtering of the Mermaid syntax before being passed to the CLI tool.

mermaid