hyperframes

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/setup.sh script installs the hyperframes CLI globally from the npm registry and uses npx puppeteer to download the chrome-headless-shell browser for high-performance video rendering.
  • [EXTERNAL_DOWNLOADS]: The npx hyperframes add command allows the agent to fetch and install additional video components, such as shader transitions and social overlays, from the hyperframes.heygen.com catalog.
  • [EXTERNAL_DOWNLOADS]: Local AI capabilities for text-to-speech (Kokoro) and transcription (Whisper) download their respective model files to a local cache folder during first-time use.
  • [COMMAND_EXECUTION]: The skill requires and executes multiple system-level CLI tools including node, npm, npx, and ffmpeg to manage project scaffolding, linting, and final video encoding.
  • [INDIRECT_PROMPT_INJECTION]: The capture functionality described in references/website-to-video.md creates a surface for indirect prompt injection by ingesting text and assets from arbitrary external URLs.
  • Ingestion points: The hyperframes capture <url> command (documented in references/cli.md) retrieves content from external websites into the local project directory.
  • Boundary markers: The provided instructions do not specify the use of delimiters or specific safety instructions to isolate captured content from the agent's primary task flow.
  • Capability inventory: The skill possesses extensive capabilities including headless browser execution (render), file system access, and network operations through the CLI.
  • Sanitization: No explicit sanitization or filtering of the captured web content is mentioned in the skill documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 06:03 AM
Security Audit — agent-trust-hub — hyperframes