The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and executes the hyperframes package from the NPM registry using npx -y hyperframes@latest during the initialization step.
[COMMAND_EXECUTION]: The skill uses shell commands to verify the environment (Node.js and FFmpeg versions) and to interact with the hyperframes CLI for project setup, linting, and video rendering.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites via WebFetch and local marketing context files. This information is used to generate video scripts and HTML templates, creating an attack surface for instructions embedded in external content.
Ingestion points: Step 1c uses WebFetch for URLs, and Step 1a reads local marketing context files.
Boundary markers: No specific delimiters or "ignore embedded instructions" markers are defined for the fetched external content.
Capability inventory: The skill utilizes Bash for rendering and Write for project file generation.
Sanitization: There is no evidence of explicit sanitization or escaping of the fetched product descriptions before they are interpolated into the beat sheet or HTML templates.
[DYNAMIC_EXECUTION]: The skill dynamically generates an index.html composition and a brand-kit.json file at runtime based on the gathered context, which are then processed by the hyperframes engine to create the video.

tiktok-launch-video