cube-build-model
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill documents the installation of the Cube CLI via a shell script fetched from the vendor's official GitHub repository (
https://raw.githubusercontent.com/cube-js/cube/master/install-cli.sh) and piped to the shell. - [EXTERNAL_DOWNLOADS]: The skill fetches the official Cube CLI installer from the vendor's public GitHub repository.
- [COMMAND_EXECUTION]: The skill utilizes multiple shell commands through the
cubeCLI to perform operations such as creating branches, forking models, and committing changes. - [PROMPT_INJECTION]: The skill processes untrusted user data (such as model definitions, SQL expressions, and branch names) which is interpolated into shell command arguments. This establishes a surface for indirect prompt injection. 1. Ingestion points: User-provided definitions for metrics, measures, and dimensions in SKILL.md. 2. Boundary markers: The skill does not provide explicit delimiters or instructions to ignore embedded commands within user content. 3. Capability inventory: The skill possesses file-writing, model modification, and deployment interaction capabilities via the
cubeCLI. 4. Sanitization: No specific instructions are provided for the agent to sanitize or validate user-supplied content before it is processed by the CLI.
Audit Metadata