The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/validate-mermaid.sh contains a code injection vulnerability. It interpolates a shell variable ($FILE) directly into a Node.js node -e command string within single quotes. If an attacker provides a filename containing a single quote and arbitrary JavaScript (e.g., test.mmd', 'utf-8'); require('child_process').execSync('whoami'); //), it could result in arbitrary code execution on the host system.
[EXTERNAL_DOWNLOADS]: The skill instructions frequently reference npx @mermaid-js/mermaid-cli and https://cdn.jsdelivr.net/npm/mermaid, which involve downloading and executing code from the NPM registry and JSDelivr CDN. While these are well-known services, they constitute external code dependencies.
[DYNAMIC_EXECUTION]: The skill provides snippets of JavaScript (for PDF generation using Puppeteer) and Python (for image processing) in the references/ directory. These are intended for runtime execution to fulfill the skill's diagramming and rendering purposes.
[PRIVILEGE_ESCALATION]: The Puppeteer configuration example in references/pdf-rendering-pipeline.md suggests using the --no-sandbox and --disable-setuid-sandbox flags. While common in containerized environments, these flags disable core browser security boundaries and increase the risk when rendering untrusted HTML content.

mermaid-diagrams