The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to download @mermaid-js/mermaid-cli from the official NPM registry if the tool is not locally available. This is a well-known service and package.
[COMMAND_EXECUTION]: The skill executes shell commands (mmdc and npx) where arguments (filenames) are derived from user-provided requirements. This creates a potential command injection vulnerability if naming conventions are bypassed during processing.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via user requirements ($ARGUMENTS). Evidence: 1. Ingestion points: User input in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Bash, Write, and Edit tools in SKILL.md. 4. Sanitization: Relies on AI adherence to naming instructions rather than programmatic filtering or escaping.

mermaid-diagram