qwen-3-5-plus-video-feedback
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The troubleshooting section in SKILL.md provides a command to install the uv tool by piping a script from https://astral.sh/uv/install.sh to the shell. This follows the official installation method for the uv tool.
- [EXTERNAL_DOWNLOADS]: The skill interacts with astral.sh for tool installation and communicates with openrouter.ai for AI inference, alongside user-configured S3-compatible storage services.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted video content which is subsequently processed by the LLM.
- Ingestion points: User-supplied video URLs or files processed by scripts/analyze_video.py.
- Boundary markers: The skill uses a structured system prompt template (assets/SYSTEM_TEMPLATE) to constrain the AI output format.
- Capability inventory: Includes file system writing via the --output parameter and network access for S3 uploads and API calls.
- Sanitization: The input is restricted by a specific system persona designed for analytical video criticism.
Audit Metadata