The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill guides the user/agent to fetch and integrate external HTML, CSS, and JavaScript snippets from a remote registry hosted on GitHub (heygen-com/hyperframes). This process lacks explicit sanitization steps in the instructions, creating a surface for indirect prompt injection where malicious code could be introduced into the local project environment if the remote source were compromised.
Ingestion points: Registry manifest (registry.json) and individual item HTML files (BLOCKNAME.html) fetched from https://raw.githubusercontent.com/heygen-com/hyperframes/main/registry in SKILL.md and references/discovery.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided for the remote content being merged into local files.
Capability inventory: The skill uses hyperframes add to write files to the local system and hyperframes render/preview which executes the fetched code in a browser environment to produce video output.
Sanitization: None specified in the workflow instructions.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from external sources, including the official HyperFrames registry on GitHub and well-known CDNs like JSDelivr for standard libraries such as GSAP and Three.js. These are documented as part of the intended functionality for building video compositions.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various CLI commands including hyperframes add, hyperframes lint, hyperframes render, and npx commands for formatting and documentation generation. It also involves running a local script scripts/generate-catalog-pages.ts as part of the contribution workflow.

hyperframes-registry