The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted data from arbitrary external URLs.
Ingestion points: Website content is extracted into files such as capture/extracted/visible-text.txt and capture/extracted/asset-descriptions.md, as described in references/step-1-capture.md.
Boundary markers: No explicit delimiters or specific instructions to ignore embedded commands within the captured content were found in the provided documentation.
Capability inventory: The skill possesses significant capabilities, including executing shell commands (npx hyperframes), writing files to the project directory, and spawning sub-agents to process the captured content.
Sanitization: There is no mention of sanitization or filtering of the extracted website text before it is used to generate the SCRIPT.md and STORYBOARD.md, creating a path where malicious site content could influence agent behavior.
[COMMAND_EXECUTION]: The workflow relies heavily on executing shell commands through the npx runner to interact with the core hyperframes toolset.
Evidence: Multiple commands are used, including npx hyperframes capture <URL>, npx hyperframes lint, npx hyperframes validate, npx hyperframes snapshot, and npx hyperframes render.
The capture command directly incorporates a user-supplied URL as an argument, which is necessary for functionality but requires the underlying tool to handle untrusted input securely.

website-to-hyperframes