qwen-mm-plugins-video-memory

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The build script script/build_memory/build_memory.sh installs necessary Python dependencies including dashscope, pydantic, numpy, and opencv-python-headless from the Aliyun PyPI mirror.\n- [COMMAND_EXECUTION]: The skill utilizes subprocess.run to execute ffmpeg and ffprobe for video frame extraction, audio chunking, and metadata retrieval. These calls use the safe list-based argument format.\n- [DATA_EXFILTRATION]: The skill communicates with official DashScope endpoints (dashscope.aliyuncs.com) for multimodal embeddings and LLM inference, and uses Alibaba Cloud OSS (oss2) for temporary storage of video clips. These are expected operations for the vendor's infrastructure.\n- [CREDENTIALS_UNSAFE]: Configuration and API keys are managed through environment variables or a local configuration file at ~/.qwen-mm-plugins/config, following standard security practices for secret management.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external video content to generate semantic memory. While malicious video content could attempt to influence the VLM's output, the skill uses structured prompts and specific task instructions to mitigate accidental obedience to embedded content.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 01:45 PM