hyperframes

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on executing shell commands for setup and core operations. The scripts/setup.sh script verifies the environment and installs tools, while the main workflow uses CLI tools like npm, npx, ffmpeg, and ffprobe to manage assets and render video.
  • [EXTERNAL_DOWNLOADS]: The installation process downloads the hyperframes CLI package from the public NPM registry and fetches a specific chrome-headless-shell binary via Puppeteer. Additionally, the skill can download further components (shader transitions and social overlays) from the vendor's catalog (hyperframes.heygen.com/catalog) at runtime.
  • [PRIVILEGE_ESCALATION]: The setup script and troubleshooting documentation suggest using sudo for global package installations (npm install -g) and system dependencies (apt-get install -y ffmpeg) if standard permissions are insufficient.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection. It is designed to ingest and process untrusted data from external URLs (using the capture command) and arbitrary text from meeting transcripts or provided assets. This data is used by the agent to generate narration scripts and HTML compositions, potentially allowing malicious content within those sources to influence the agent's behavior.
  • Ingestion points: External websites (via hyperframes capture), meeting transcripts, and cloud-stored assets (Google Drive/Adobe Library).
  • Boundary markers: None explicitly implemented to isolate extracted web text or transcript content from the agent's instructional context.
  • Capability inventory: The skill can execute shell commands (npm, ffmpeg), write local files (HTML, JS, shell scripts), and download external code.
  • Sanitization: No specific sanitization or filtering of extracted text is described before it is processed by the agent.
  • [DYNAMIC_EXECUTION]: The skill's primary function involves dynamically generating HTML, CSS, and GSAP-based JavaScript code which is then executed within a headless browser environment to render video frames. It also includes a helper script (animation-map.mjs) for analyzing these generated timelines.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 07:35 PM
Security Audit — agent-trust-hub — hyperframes