youtube-transcribe

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

• Third-party content exposure detected (high risk: 0.85). The skill ingests the video’s audio transcript by running transcribe.py with OpenAI Whisper and then reads/writes the resulting result["text"] into youtube-transcribe/transcript.txt/transcript.json, so any free-form text content originates from an outsider-authored video file provided by the user (not authored by the operating user) and becomes LLM-readable context in Step 5/6.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (medium risk: 0.60). The prompt mainly writes project-local files and runs user-level Python scripts (no sudo, no user creation, no systemctl/ssh edits), but it includes an explicit SSL certificate-verification bypass (ssl._create_default_https_context = ssl._create_unverified_context), which is a security-risking instruction.