caveman

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No security threats detected. The skill focuses exclusively on modifying the AI's linguistic style to be more concise.
  • [PROMPT_INJECTION]: No attempts to override system prompts or bypass safety filters were found. The 'caveman' mode is a formatting instruction triggered by explicit user request.
  • [DATA_EXFILTRATION]: There are no network requests, file system access, or credential handling operations.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute any external scripts or packages.
  • [INDIRECT_PROMPT_INJECTION]: While the skill instructions dictate how to handle user queries, it specifically includes an 'Auto-Clarity Exception' that prioritizes security warnings and clarity over the persona, reducing risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 10:39 AM
Security Audit — agent-trust-hub — caveman