ai-video-generation

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). These URLs are suspicious because they come from an untrusted/unknown domain and the skill explicitly instructs piping a remote install script (curl -fsSL https://cli.inference.sh | sh), a high‑risk pattern for executing attacker‑controlled code, and they include direct downloadable endpoints (media/files/hosted images) and placeholder/ambiguous hostnames that could be abused to deliver or hide malicious payloads.

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md explicitly shows runtime inputs that fetch and process arbitrary external URLs (e.g., Image-to-Video and Avatar examples using "image_url": "https://your-image.jpg", "audio_url": "https://speech.mp3", and video_url examples), so the agent will ingest untrusted, user-provided web content that could contain instructions or signals influencing model outputs and subsequent actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The Quick Start installs and executes remote code at runtime via "curl -fsSL https://cli.inference.sh | sh", which fetches and runs a script (https://cli.inference.sh) that is a required dependency for the skill and therefore can directly execute remote code.