cube-explore-model
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill includes a setup command (
curl -fsSL ... | sh) that downloads and runs an installation script for the Cube CLI. The script is hosted on the Cube organization's official GitHub repository, matching the skill's vendor identity. - [COMMAND_EXECUTION]: The skill uses the
cubeCLI to perform data model inspection tasks such as listing deployments, fetching file contents, and querying metadata. All documented commands are read-only and intended for exploration. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data from data model files and compiled metadata, creating a surface for indirect prompt injection.
- Ingestion points: Data model files (.yml) and compiled metadata retrieved via the
cube data-model getandcube metacommands (SKILL.md). - Boundary markers: Absent; the instructions do not specify delimiters for the retrieved data content.
- Capability inventory: The skill's capabilities are focused on metadata inspection and retrieval via the CLI; it does not perform arbitrary file writes or external network operations beyond the tool's scope.
- Sanitization: Absent; the agent receives raw content from the semantic model files.
Audit Metadata