The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes npx to download and execute pinned versions of @mermaid-js/mermaid-cli@11.17.0 and @probelabs/maid@0.0.29 from the NPM registry. These are well-known services, and the versions are explicitly pinned to ensure consistency and security.
[COMMAND_EXECUTION]: The render.py, mermaid_lint.py, and svg_lint.py scripts use subprocess.run() to call various system binaries such as mmdc, rsvg-convert, dot, and headless Chrome. The command construction uses list-based arguments, which mitigates the risk of shell-based command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted diagram source code provided by users, creating a surface for potential indirect prompt injection.
Ingestion points: Diagram source files including .mmd, .svg, and .dot, as well as Mermaid blocks within Markdown files.
Boundary markers: The instructions explicitly require running deterministic lints and a security gate before rendering any content.
Capability inventory: The skill has the ability to write rendered PNG files to the local file system and execute CLI tools.
Sanitization: The svg_lint.py script implements a strict whitelist for SVG elements and attributes. It explicitly detects and blocks active content, including <script> tags, event handlers (e.g., onclick), and external href references to prevent cross-site scripting (XSS) or renderer-based exploits.

diagram-review