video-style-analysis

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This skill explicitly implements high-fidelity video/audio cloning (face detection + cross-frame identity clustering, voice cloning/TTS, image/video generation via third‑party APIs) and uploads user media to external providers, enabling easy impersonation and privacy-invasive deepfakes — a high abuse risk even though there are no hidden eval/remote‑exec backdoors or covert credential‑exfiltration routines in the code.

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

• Third-party content exposure detected (high risk: 0.85). 步骤3.5/3.6/4/5/6/7由 Claude 读取并处理 output/analysis/asr_result.json 的 full_text（Whisper 从“原视频”的音频转写得到的解说文本），该文本属于操作用户未选择引入的“原视频内容/第三方创作者的口播/解说”自由文本，因此会进入 LLM 上下文；路径：asr_transcriber.py → output/analysis/asr_result.json → Claude 读取 full_text/segments。