docker-ai-ml
Warn
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires administrative privileges to perform system-level configuration, specifically for installing the NVIDIA Container Toolkit and managing background services (e.g., sudo apt-get install -y nvidia-container-toolkit and sudo systemctl restart docker).
- [COMMAND_EXECUTION]: The skill utilizes docker run, docker exec, and the docker agent CLI to execute commands. The agent.yaml configuration explicitly grants agents the ability to execute_command and write_file on the host or within sandboxes.
- [EXTERNAL_DOWNLOADS]: Fetches cryptographic keys and repository configuration files from NVIDIA's official GitHub-hosted domain (nvidia.github.io) to enable GPU support.
- [EXTERNAL_DOWNLOADS]: The docker model skills command involves fetching and installing additional capabilities for AI tools such as Claude Code and Codex from remote sources.
- [PROMPT_INJECTION]: The multi-agent orchestration feature (docker agent run) presents a surface for indirect prompt injection by processing user prompts and local files with powerful tools. * Ingestion points: User instructions passed to the CLI and configuration defined in agent.yaml (SKILL.md). * Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands in processed data. * Capability inventory: Access to write_file, execute_command, and web_search (SKILL.md). * Sanitization: No evidence of input validation or sanitization before passing data to the agent orchestration layer.
Audit Metadata