The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent on how to use interactive features of Mermaid, such as click events and tooltips (detailed in references/advanced-features.md). If an agent generates a diagram based on untrusted user input and the platform renders it, these interactive elements could be used for phishing or directing users to malicious sites. • Ingestion points: User-provided technical descriptions or requests to "diagram" or "visualize" (defined in SKILL.md). • Boundary markers: The skill uses standard markdown code blocks (mermaid) to delimit diagram content. • Capability inventory: The skill teaches the agent to generate diagrams that can include external links (click command) and custom styling. • Sanitization: The skill does not provide instructions for sanitizing user-provided text before incorporating it into diagram labels or links.
[EXTERNAL_DOWNLOADS]: The skill documentation refers to standard external resources for rendering and extending Mermaid diagrams. • Evidence: Instructions in SKILL.md and README.md for installing @mermaid-js/mermaid-cli via npm. • Evidence: References in references/architecture-diagrams.md for installing @iconify-json/logos and other icon packs. • Evidence: Example HTML integration in references/advanced-features.md using the jsdelivr.net CDN to load the Mermaid library.

mermaid-diagrams