mk-youtube-search

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill automatically downloads the jq binary from the official jqlang/jq GitHub repository during initialization if not found in the system path or local bin directory.
  • [EXTERNAL_DOWNLOADS]: The skill automatically downloads the yt-dlp binary from the official yt-dlp/yt-dlp GitHub repository if the tool is not available locally.
  • [COMMAND_EXECUTION]: The main script executes the yt-dlp and jq binaries to perform YouTube searches and transform the results into structured JSON.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data (video titles and descriptions) fetched from YouTube search results.
  • Ingestion points: Search metadata returned by yt-dlp in scripts/search.sh.
  • Boundary markers: Absent; data is parsed into JSON and returned to the agent context without explicit delimiters or warnings to ignore embedded instructions.
  • Capability inventory: Subprocess execution of yt-dlp and jq, and file writing to the /tmp/monkey_knowledge/ directory.
  • Sanitization: The skill uses jq to structure the data, but it does not sanitize or filter the text content of title or description fields for potential injection patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 12:00 PM