The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to generate HTML files that load CSS and JavaScript assets from the JSDelivr CDN (cdn.jsdelivr.net). These assets, specifically Tailwind CSS and Highlight.js, are used to provide consistent styling and syntax highlighting across the course material. JSDelivr is a widely recognized and reputable service for delivering open-source assets.
[COMMAND_EXECUTION]: The skill instructions suggest that the user or agent can serve the workspace directory using the standard python -m http.server command. It also notes that the agent should open lesson files for the user via a CLI command if the environment supports it. these are standard development tasks and do not represent malicious activity.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it incorporates external resources and user-provided definitions into its generated lessons. While this involves processing untrusted data, the agent's role as an orchestrator of educational content and the structured nature of the files mitigate the risk.
Ingestion points: RESOURCES.md (external links and descriptions) and GLOSSARY.md (user-provided terminology).
Boundary markers: No explicit delimiters or warnings are defined for the interpolation of external data into the generated lessons.
Capability inventory: The skill performs local filesystem writes (HTML, Markdown) and suggests CLI commands to the user.
Sanitization: No specific sanitization or filtering logic is prescribed for the ingested content.

teach