The Agent Skills Directory

[COMMAND_EXECUTION]: The render.sh script executes the mmdc (Mermaid CLI) and imgcat commands. The script uses Bash arrays (MMDC_ARGS) to handle parameters, which prevents shell injection vulnerabilities by ensuring arguments are not parsed as commands.
[EXTERNAL_DOWNLOADS]: The skill references the @mermaid-js/mermaid-cli Node.js package. This is a well-known technology used for diagram rendering and is considered a trusted source.
[INDIRECT_PROMPT_INJECTION]: The skill processes diagram definitions stored in .mmd files. It includes the following security context:
Ingestion points: Reads diagram content from files (e.g., /tmp/mermaid-diagram.mmd or user-specified paths) in render.sh and SKILL.md.
Boundary markers: None explicitly defined for the diagram content itself.
Capability inventory: The skill can write files, execute the render.sh script, and run mmdc via Bash.
Sanitization: Command-line arguments are sanitized via array interpolation in the shell script, limiting the exploitability of the processing pipeline.

mermaid-render