media-fetcher

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The capture_screenshot.py script automatically installs the playwright Python package from PyPI and the Chromium browser from Microsoft's official servers if they are not detected on the system. These downloads originate from well-known, trusted technology providers.
  • [COMMAND_EXECUTION]: The skill's core functionality relies on executing external command-line utilities via the subprocess module.
  • scripts/capture_screenshot.py executes ffprobe to extract video stream dimensions.
  • scripts/download_video.py invokes yt-dlp to download video content from remote URLs.
  • scripts/render_diagram.py calls the Mermaid CLI (mmdc) to render diagrams to image files.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface (Category 8) because it processes untrusted data from segments.json to drive its actions.
  • Ingestion points: The skill reads segments.json (generated from video transcriptions) to extract brand names, technical concepts, and URLs.
  • Boundary markers: No explicit delimiters or instructions are used to separate the ingested data from the agent's internal logic.
  • Capability inventory: The skill possesses capabilities for browser automation (Playwright), network downloads (yt-dlp, urllib), and local file system writes.
  • Sanitization: Although download_video.py sanitizes output filenames, the search queries and URLs extracted from the metadata are processed directly, which could lead to navigating to malicious sites or downloading unintended content if the transcription data is manipulated.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 02:27 AM