youtube-live-chat
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-generated content from YouTube live chat, representing a surface for potential indirect prompt injection attacks. * Ingestion points: The parse_chat_json function in get_live_chat.py reads JSON data containing messages from the external YouTube service. * Boundary markers: None; the script parses and displays raw message strings without delimiters or instructions to ignore embedded commands. * Capability inventory: The skill can write the resulting transcript to a user-specified file system path or output it to the terminal. * Sanitization: None; the script extracts message text directly from the JSON payload without filtering or escaping.
- [EXTERNAL_DOWNLOADS]: The skill downloads chat metadata from YouTube using the well-known yt-dlp library. This connection to a well-known service is safe and essential for the functionality of the skill.
Audit Metadata