The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites via the hyperframes capture tool. This content is used to inform design decisions and generate video scripts. A malicious website could contain hidden instructions designed to manipulate the agent's behavior or output.
Ingestion points: capture/extracted/visible-text.txt and capture/extracted/asset-descriptions.md containing data from captured websites.
Boundary markers: The skill does not define explicit delimiters or instructions to ignore potential injections within the captured text.
Capability inventory: The skill can execute shell commands, write files to the local system, and generate executable HTML/JS compositions.
Sanitization: There is no mention of sanitizing or escaping the text extracted from websites before it is used in prompt generation or script authoring.
[COMMAND_EXECUTION]: The skill uses npx hyperframes and curl to manage the video production lifecycle, including site capture, linting, rendering, and TTS generation. These commands often take external inputs such as URLs.
[EXTERNAL_DOWNLOADS]: The workflow involves downloading assets from external websites and audio files from third-party TTS providers (HeyGen, ElevenLabs).
[DYNAMIC_EXECUTION]: The skill generates HTML and JavaScript composition files for each video 'beat'. While these are based on templates, they incorporate dynamic content and styles derived from the untrusted external website data.

website-to-video