grok-cli-runtime
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
grokCLI binary through a shell script to verify service availability and manage session states. - [CREDENTIALS_UNSAFE]: The skill manages the
~/.grok/auth.jsoncredential file and handlesGROK_AUTH_B64environment variables. It provides explicit instructions to treat these files as opaque to prevent data exposure during recovery and deposition processes. - [INDIRECT_PROMPT_INJECTION]: The skill parses model output from the
grokcommand to diagnose service failures, which presents an attack surface where model output influences agent actions. - Ingestion points: Output from the diagnostic
grokprobe inSKILL.md. - Boundary markers: The instructions specify checking specific JSON structural fields (
.type,.stopReason) to identify errors, which helps delimit data from instructions. - Capability inventory: The agent has the capability to execute shell commands (
grok) and write sensitive data to~/.grok/auth.json. - Sanitization: The skill requires structural validation of the JSON response (checking for
errortype and exit codes) before processing the message content.
Audit Metadata