The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The documentation describes how to include interactive elements and external resources in diagrams, creating a surface for injection if an agent renders untrusted content.\n
Ingestion points: Mermaid syntax blocks within the documentation and potentially user-provided data during agent operation.\n
Boundary markers: None identified in the skill instructions to separate trusted and untrusted diagram components.\n
Capability inventory: The skill supports click events that execute JavaScript callbacks and redirection to external URLs, as well as embedding remote images.\n
Sanitization: The skill mentions that these interactive features are subject to the securityLevel configuration, which defaults to 'strict' to prevent execution.\n- [DYNAMIC_EXECUTION]: The skill provides examples of using the click command to execute JavaScript functions (call callback()) from within diagram nodes when interactive features are enabled.\n- [EXTERNAL_DOWNLOADS]: Demonstrates how to load external assets, including images from example.com and mermaid.js.org, and integrating the FontAwesome library via a Cloudflare-hosted CDN.

mermaid-flowchart