skills/varghq/skills/varg-previz/Gen Agent Trust Hub

varg-previz

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill implements a legitimate local development environment for video planning. It contains a local HTTP server (serve.py) to host a three.js application, which is a standard pattern for interactive AI agent tools.
  • [COMMAND_EXECUTION]: The skill utilizes subprocess.run in build_pack.py to execute ffmpeg for video encoding and concatenation. These calls use argument lists rather than raw shell strings, which is a safe practice that prevents command injection. The usage of FFmpeg is central to the skill's purpose of building video reference packs.
  • [DATA_EXPOSURE]: The local server in serve.py implements a _local() security check. This check validates that the Host header is localhost or 127.0.0.1 and ensures that any Origin header matches the local server, preventing Cross-Site Request Forgery (CSRF) or unauthorized remote access to the local save/export APIs.
  • [FILE_SYSTEM]: The /api/export and /api/save endpoints in the local server are restricted. The export function specifically enforces that only files ending in .md can be written to the project directory, and it uses the .name attribute to prevent path traversal attacks.
  • [EXTERNAL_DOWNLOADS]: The skill references standard three.js libraries and utilizes Playwright for headless browser frame capture. All external service references (e.g., varg.ai, s3.varg.ai) are consistent with the vendor's own infrastructure and the primary purpose of the skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 09:50 PM