The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and 'port' existing web pages, including their DOM, CSS, and JavaScript, to create slideshow compositions. This introduces a risk surface where instructions or malicious logic in the source content could be processed by the agent.
Ingestion points: External source pages and their associated mechanical visuals (referenced in the 'Porting source pages' section).
Boundary markers: There are no instructions to wrap external content in delimiters or use safety-guarding prompts when processing source data.
Capability inventory: The agent generates HTML and JavaScript (GSAP timelines) and can execute local linting tools.
Sanitization: No specific sanitization, filtering, or escaping logic is provided for handling untrusted scripts from source pages.
[COMMAND_EXECUTION]: The skill documentation includes instructions for running npx hyperframes lint to validate the composition's logic and structure.
[EXTERNAL_DOWNLOADS]: The standalone harness reference implementation downloads the Three.js library from the JSDelivr CDN, which is a well-known service.

slideshow