rbac-design

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides a comprehensive framework for RBAC/ABAC design based on NIST standards and industry best practices. It does not perform any sensitive operations or maintain persistence.
  • [PROMPT_INJECTION]: The skill contains defensive instructions intended to harden the agent against adversarial content in role definitions. A static detection for 'ignore previous instructions' was evaluated as a false positive, as the phrase is used within a security boundary section instructing the agent to ignore such injections found in processed data.
  • [DATA_EXFILTRATION]: The skill's capabilities are restricted to local read-only operations via the 'allowed-tools' (Read, Grep, Glob). There are no tools for network communication or external data transfer, and no evidence of credential harvesting was found.
  • [COMMAND_EXECUTION]: No shell commands, script generation, or dynamic code execution patterns were detected. The skill is designed for architectural review and recommendation only.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 01:36 PM
Security Audit — agent-trust-hub — rbac-design