short-form-video-clone-edit

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This is high-risk: the skill is an explicit, targeted non‑consensual deepfake/impersonation pipeline (names a specific influencer), including voice cloning, exhaustive frame scanning to capture fleeting appearances, instructions to strip original audio, and use of external APIs plus stored API keys/ direct DB connection strings—facilitating deliberate abuse—while I found no obfuscated payloads, remote shells, or explicit credential-exfiltration/backdoor code in the text itself.

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly downloads public influencer videos (Step 1: "yt-dlp" from VIDEO_URL, e.g., Instagram @nick_saraev) and then transcribes, analyzes frames (face detection), and uses that untrusted third-party content to decide segmenting and avatar insertion, so external page/user-generated content directly influences tool actions.