rbac-design

Pass

Audited by Gen Agent Trust Hub on May 8, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides architectural patterns and checklists for RBAC and ABAC design without including executable code or dangerous system commands.
  • [SAFE]: It utilizes platform-level restrictions via the allowed-tools field to limit its scope to benign operations like reading and searching files.
  • [PROMPT_INJECTION]: Phrases like 'ignore previous instructions' were detected but found to be part of an 'Injection Hardening' section specifically designed to protect the agent from adversarial content in input files.
  • [SAFE]: The skill addresses potential indirect prompt injection by providing a clear evidence chain: it ingests untrusted design files via arguments, defines a security boundary in the 'Injection Hardening' section, limits capabilities to Read, Grep, and Glob tools, and instructs the agent to treat input as untrusted and flag suspicious directives.
Audit Metadata
Risk Level
SAFE
Analyzed
May 8, 2026, 12:28 AM
Security Audit — agent-trust-hub — rbac-design