skills/varghq/skills/threejs-animatic/Gen Agent Trust Hub

threejs-animatic

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches the three.js library and its addons from jsDelivr (a well-known CDN) using an import map in index.html. This is a standard practice for web development and provides a pinned, verifiable version of the library.
  • [COMMAND_EXECUTION]: The starter/tools/ scripts utilize Bun.spawn to execute local commands including ffmpeg for video encoding and google-chrome (or chromium) for headless rendering. These operations are restricted to the local environment and are essential for the skill's stated purpose of generating MP4 animatics.
  • [DATA_EXPOSURE]: The local development server in starter/serve.ts is configured to listen only on 127.0.0.1. It includes restricted API endpoints for saving project state to data/edits.json and exporting camera plans to specific markdown and JSON files, with path normalization to prevent directory traversal.
  • [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves the agent ingesting user-provided brief.md files to generate Three.js code.
  • Ingestion points: brief.md and project-specific descriptions in js/shots.js.
  • Boundary markers: The agent is instructed to follow specific schemas and code structures.
  • Capability inventory: The system can write to local files (data/edits.json) and execute local rendering tools (ffmpeg, chrome).
  • Sanitization: The skill relies on the agent's interpretation and standard guardrails when transforming natural language briefs into executable Three.js code.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 09:49 PM