youtube-edit-kit

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow ingests outsider-authored free text via the user-provided source video’s audio, which is transcribed at G1 (“Transcribe ONCE… READ the entire packed transcript”), and that transcript text then drives later caption-term review and corrections.

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

• Potentially malicious external URL detected (high risk: 0.90). The README and llms.txt explicitly instruct agents to git clone https://github.com/crealwork/youtube-edit-kit or to fetch multiple raw.githubusercontent.com URLs (for example https://raw.githubusercontent.com/crealwork/youtube-edit-kit/main/SKILL.md and BOOTSTRAP.md) at runtime to load the skill docs/prompts and runnable scripts, which directly control agent behavior and execution.