docker-ai-ml
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill contains instructions to fetch GPG keys and repository configuration files from NVIDIA's official GitHub infrastructure. These downloads are a standard part of installing the NVIDIA Container Toolkit for GPU support.
- [PRIVILEGE_ESCALATION]: Guidance for setting up GPU acceleration on Linux includes the use of
sudofor package installation and service management. These actions are required for system-level configuration of hardware passthrough. - [INDIRECT_PROMPT_INJECTION]: The skill documents the creation of Docker Agents with tools such as
execute_commandandwrite_file. This defines a vulnerability surface where the agent could be manipulated if it processes untrusted data without proper sanitization. - Ingestion points: The agent's task input and configuration files like
agent.yamlinSKILL.md. - Boundary markers: The examples do not include explicit delimiter markers or instructions to ignore embedded commands.
- Capability inventory: The agent is granted capabilities to execute shell commands and modify the filesystem within its container as described in
SKILL.md. - Sanitization: No sanitization or validation logic is present in the architectural examples provided.
Audit Metadata