youtube-summary

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data from YouTube transcripts and metadata, creating a surface for indirect prompt injection.\n
  • Ingestion points: Video metadata (title) and transcript segments ingested via scripts/fetch.py and scripts/segments.py.\n
  • Boundary markers: Absent in transcript interpolation, though the prompts (prompts/extract_structure.md, prompts/summarize_sections.md) mandate JSON output which provides structural separation.\n
  • Capability inventory: Subprocess execution of local scripts and file system writes to ~/youtube-reports/.\n
  • Sanitization: Filenames are slugged in scripts/fetch.py using _slug() to prevent path traversal via video titles.\n- [COMMAND_EXECUTION]: The skill uses subprocess calls to execute its own local Python scripts (fetch.py, cache.py, segments.py, doctor.py). These scripts perform data processing, caching, and environment validation tasks.\n- [EXTERNAL_DOWNLOADS]: The skill downloads YouTube transcripts and metadata using well-known libraries (yt-dlp, youtube-transcript-api). It also fetches thumbnails from official YouTube/Google domains. All external network activity targets well-known services.\n- [SAFE]: No evidence of hardcoded credentials, malicious obfuscation, or persistence mechanisms was found. The use of __import__ in scripts/doctor.py is restricted to a hardcoded list of required packages for dependency checking.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 12:10 PM
Security Audit — agent-trust-hub — youtube-summary