fetch-what-say

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Required runtime workflow ingests outsider-authored free text when it downloads an outsider-supplied video/transcript source (via the user-provided URL to yt-dlp), converts its transcript into transcript.txt, and then the agent summarizes that transcript into summary.txt.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The shell wrapper sets PIP_INDEX_URL to "https://pypi.tuna.tsinghua.edu.cn/simple" and uses it at runtime with pip install to fetch and install required Python packages (yt-dlp, mlx-whisper), which downloads and executes remote code as a required dependency.