workbench-self-awareness-infra

Pass

Audited by Gen Agent Trust Hub on May 3, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides structural guidance for agent behavior and environment verification without including executable code or shell commands.
  • [SAFE]: It contains explicit safety rules in the 'Field Rules' and 'Anti-Patterns' sections that prohibit the exposure of private IDs, tokens, direct IPs, environment dumps, cookies, or request payloads.
  • [SAFE]: The instructions encourage best practices for risk management, such as defining a 'risk_envelope' and establishing clear 'role_boundary' limits for agent autonomy.
Audit Metadata
Risk Level
SAFE
Analyzed
May 3, 2026, 03:40 AM