cube-configure-agent
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill mentions an installation command for the Cube CLI that pipes a remote script to the shell:
curl -fsSL https://raw.githubusercontent.com/cube-js/cube/master/install-cli.sh | sh. This script is hosted on the official GitHub repository of the developer, cube-js. - [EXTERNAL_DOWNLOADS]: Downloads an installation script from the cube-js GitHub organization.
- [COMMAND_EXECUTION]: Provides examples for executing the
cubeCLI to manage the semantic model, specifically for reading and writing agent configuration files. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for managing the agent's core instructions, which are stored as markdown files and processed at runtime. 1. Ingestion points: Configuration is read from the
agents/directory usingcube data-model getandcube:readDataModelFile. 2. Boundary markers: Not explicitly defined in the prompt instructions. 3. Capability inventory: The agent has the ability to write to the data model (cube data-model put), commit changes (cube:commitDataModelChanges), and list cloud deployments. 4. Sanitization: The instructions state that these files inherit the project's git flow, review, and access policies, providing a human-in-the-loop review mechanism for new rules and skills.
Audit Metadata