website-to-video

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites (captured as visible-text.txt and asset-descriptions.md) which is used to generate summaries and storyboards. This presents a potential surface for indirect prompt injection if the website content contains malicious instructions designed to influence the agent's behavior during the design or briefing phases.
  • Ingestion points: capture/extracted/visible-text.txt and capture/extracted/asset-descriptions.md are read during Steps 2 and 3.
  • Boundary markers: The instructions do not explicitly mandate the use of delimiters or warnings to ignore embedded instructions in the captured site text.
  • Capability inventory: The skill has capabilities to execute shell commands (npx hyperframes, curl, node), write files to the project directory, and perform network operations to AI service providers.
  • Sanitization: There is no evidence of explicit sanitization or filtering of the captured site text before it is interpolated into prompts for design and strategy.
  • [COMMAND_EXECUTION]: The skill uses various shell commands to perform its duties, including npx hyperframes for site capture and video rendering, curl for communication with HeyGen and ElevenLabs APIs, and node to run project verification scripts. These commands are integral to the skill's documented purpose and target legitimate tools and endpoints.
  • [DYNAMIC_EXECUTION]: The workflow involves generating HTML and JavaScript files (index.html and individual beat compositions) that incorporate project-specific logic and brand styles. These generated scripts are then executed within a headless Chrome environment via the HyperFrames CLI to generate snapshots and final video renders.
  • [EXTERNAL_DOWNLOADS]: The skill references and fetches resources from external sources, such as fonts from the target website, sound effects from Pixabay, and JavaScript animation libraries (GSAP, Three.js) from well-known CDNs like JSDelivr for use in the video compositions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 06:32 PM