website-to-hyperframes
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the
npx hyperframesCLI tool to execute various tasks including website capture, linting, validation, and rendering. - [EXTERNAL_DOWNLOADS]: It downloads screenshots, assets, and design tokens from user-specified external URLs during the initial capture phase.
- [REMOTE_CODE_EXECUTION]: Generated video compositions reference external scripts from the jsdelivr CDN, such as GSAP for animations and the DotLottie player for vector graphics.
- [PROMPT_INJECTION]: The ingestion of untrusted content from external websites represents an indirect prompt injection vector.
- Ingestion points: External website text and metadata are saved to the
capture/directory for analysis (e.g., in capture/extracted/visible-text.txt). - Boundary markers: No specific delimiters are used to separate untrusted content from instructions.
- Capability inventory: The environment allows for subprocess execution via
npx, file system modifications, and network access. - Sanitization: Instructions include basic filtering of HTML tags from extracted text before it is used in script writing.
Audit Metadata