read-youtube-feed
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to execute a local Python script
scripts/get_updates.py. This script is used to scrape YouTube channel pages and process video metadata. - [EXTERNAL_DOWNLOADS]: The Python script performs network operations to fetch data from
www.youtube.comandr.jina.ai(a well-known service for LLM-friendly web scraping) to retrieve video information and descriptions. - [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from the web and presents it to the agent.
- Ingestion points: Untrusted data (video titles and descriptions) is ingested via
requests.getinscripts/get_updates.py(lines 80, 168). - Boundary markers: The skill lacks explicit delimiters or instructions to the agent to ignore any potential commands embedded within the retrieved video descriptions.
- Capability inventory: The skill performs network requests and is part of a larger workflow described in
SKILL.mdthat involves processing content and saving data to external platforms like Feishu/Lark. - Sanitization: While
generate_summary(line 228) filters out some formatting and links, it does not sanitize the text for potentially malicious instructional content.
Audit Metadata