caveman

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The instructions in SKILL.md define a persona for token efficiency ('caveman mode') which drops filler words while maintaining technical substance. These are standard behavioral instructions and do not attempt to bypass safety filters.
  • [SAFE]: The skill includes an 'Auto-Clarity Exception' rule that explicitly directs the agent to drop the persona for security warnings and irreversible actions, which is a defensive best practice.
  • [SAFE]: The provided Python scripts (scripts/caveman_compressor.py, scripts/caveman_lint.py, and scripts/token_savings_estimator.py) are deterministic, regex-based utilities. They rely solely on the Python standard library, perform no network operations, and do not execute external code or commands.
  • [SAFE]: References to external repositories (e.g., Matt Pocock's GitHub) are used for attribution and are documented clearly without obfuscation.
  • [SAFE]: No hardcoded credentials, sensitive file access, or persistence mechanisms were found in any of the skill files or scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 12:25 AM
Security Audit — agent-trust-hub — caveman