herdr-grok-review

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The documents explicitly instruct launching an agent with sandboxing disabled and full/trust/force flags and to run arbitrary shell probes, which intentionally bypasses safety controls and enables remote code execution and data-exfiltration abuse.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The prompt explicitly instructs launching the reviewer with sandboxing disabled and workspace trust/“full permissions” (e.g. "--sandbox disabled --trust" and approve flags), which bypasses security controls and could enable the agent to modify the host environment.