screenkite-clean-cut

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

• Third-party content exposure detected (high risk: 0.80). The skill's workflow (SKILL.md Phase 2 and the transcribe_mic.py → compute_all_cuts.py scripts) ingests word-level JSON transcripts produced by the external ElevenLabs Scribe service from mic audio and then parses that untrusted, user-generated transcript text to compute editTimeline cut ranges that directly drive tool calls, so third-party content can influence actions.