grok-coding-agent
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the execution of the grok CLI tool and user-provided wrapper scripts. It includes instructions for capability discovery using --version and --help flags.
- [DATA_EXFILTRATION]: The skill includes explicit security precautions to prevent data exposure, instructing the agent to never dump environment variables or wrapper source code which might contain credentials.
- [PROMPT_INJECTION]: The skill acts as a proxy for user-provided tasks, creating an indirect prompt injection surface.
- Ingestion points: User instructions are passed to the external executor via the -p parameter.
- Boundary markers: No explicit delimiters are used; the skill relies on the tool's standard argument parsing.
- Capability inventory: The external tool can perform file system operations (workspace edits) and code analysis when the workspace sandbox is selected.
- Sanitization: The skill relies on the inherent sandboxing features of the grok tool (e.g., read-only and workspace sandboxes) to limit the impact of potentially malicious inputs.
Audit Metadata