The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's HTML template (assets/word_card.html) fetches the Tailwind CSS library and Mermaid.js library from the JSDelivr content delivery network.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection where malicious input could lead to script execution in the generated HTML file.
Ingestion points: The word provided by the user is the primary ingestion point, which is then used to generate definitions, examples, and Mermaid charts inside SKILL.md.
Boundary markers: No boundary markers or instructions to sanitize user input are present in the skill instructions.
Capability inventory: The skill instructions specify writing generated content to local HTML files (e.g., word_card_{word}.html).
Sanitization: There is no evidence of sanitization or escaping of the user-provided word or the LLM-generated content before it is written to the HTML file.
Evidence: The file assets/word_card.html contains a Mermaid initialization block with securityLevel: 'loose', which explicitly disables some of Mermaid's built-in XSS protections, making the generated file more vulnerable to malicious input injected via the word deconstruction process.

explain-words