container-ops
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill describes commands that fetch container images from Docker Hub, such as nginx and ubuntu. This is standard behavior for container management tools.
- [COMMAND_EXECUTION]: The skill instructs the agent to use shell command substitution for bulk management tasks, such as stopping or removing all containers via $(docker ps -q), which increases the potential impact of unintended command execution.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates executing Docker commands using external parameters without defining safety boundaries.
- Ingestion points: User input is expected for container names, environment variables, volume mappings, and command strings within the docker exec context.
- Boundary markers: No specific delimiters or warnings are included to ensure the agent treats user input as non-executable data.
- Capability inventory: Grants the agent significant control over container lifecycles and file transfers between the host and containers.
- Sanitization: No input validation or sanitization logic is provided to mitigate command injection risks.
Audit Metadata