The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill manages the installation of specialized video workflows by downloading additional skills from the HeyGen repository using the npx skills add command.
[REMOTE_CODE_EXECUTION]: To maintain current versions, the skill uses npx hyperframes init and npx hyperframes skills update to fetch and run synchronization logic from remote sources.
[COMMAND_EXECUTION]: The framework relies on several command-line tools to function, including npx for package management, the GitHub CLI (gh) for reading repository data, and the hyperframes utility for core rendering operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external sources, presenting a risk for indirect prompt injection.
Ingestion points: The skill is designed to crawl product and website URLs using headless Chrome, ingest GitHub Pull Request data via the gh CLI, and process user-supplied video and audio files for transcription and captioning.
Boundary markers: No specific delimiters or safety warnings are specified to help the agent differentiate between the developer's instructions and untrusted content retrieved from external sources.
Capability inventory: Successful exploitation could leverage the skill's access to shell commands (npx, gh), network capabilities (crawling), and the ability to read and write to the local file system during video composition and rendering.
Sanitization: There is no mention of sanitizing or validating the data retrieved from external URLs, GitHub repositories, or media files before it is incorporated into the agent's context.

hyperframes