video-asset-sources

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses curl to fetch assets and metadata from a wide array of third-party domains including Pexels, Pixabay, Freesound, Mixkit, Coverr, Wikimedia, NASA, and the Internet Archive. While these are mostly established services, the sheer volume of external network requests increases the skill's attack surface.\n- [COMMAND_EXECUTION]: The skill instructs the agent to run remote code via npx for packages like hyperframes (a vendor-associated tool) and remotion-bits (a third-party utility). These commands download and execute code directly from the npm registry at runtime.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the web to extract URLs and licensing information. This creates a surface for tool output poisoning where malicious site content could influence agent behavior.\n
  • Ingestion points: curl fetching from various external websites (Mixkit, Coverr, etc.) to grep for media URLs.\n
  • Boundary markers: Absent for the scraped content; the agent is expected to process the raw output of the search results.\n
  • Capability inventory: curl (network), Write (file system), python3 (local script execution), npx (remote package execution).\n
  • Sanitization: The record_asset.py script uses regex to filter license strings, providing validation for that specific field, but does not sanitize other ingested metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 09:49 PM