skills/microsoft/hve-core/caveman/Gen Agent Trust Hub

caveman

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFE
Full Analysis
  • Instructional Output Formatting: The skill defines rules for a terse 'Caveman' response style. These are standard behavioral guidelines and do not include attempts to override safety protocols or bypass system instructions.
  • Security-First Clarity Rules: The skill explicitly includes boundaries where compression is disabled, such as when communicating vulnerability disclosures or confirming destructive actions. This preserves the integrity of safety-critical communications.
  • No External Capabilities: The skill does not perform network requests, access the file system, or execute code. It relies solely on the conversation transcript for state, minimizing its attack surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 06:10 PM
Security Audit — agent-trust-hub — caveman