moai-ref-cross-model-audit
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines instructions for the
mcp__moai__audit_multitool, which is a specialized component of the modu-ai platform designed for cross-model convergence in code reviews. The behavior described involves parallel processing of audit tasks across multiple backends (Claude, Codex, and GLM). - [SAFE]: The "Independence rule" documented in the skill is a security and quality best practice designed to prevent model bias. It instructs the agent to provide only high-level metadata to secondary backends rather than full analysis text, ensuring uncorrelated second opinions.
- [SAFE]: Data persistence mentioned (storing results in
.moai/state/audit-multi/) follows standard application state management practices for this agent's domain. - [SAFE]: The convergence logic (pass/fail/disagreement handling) is a transparent policy for synthesizing multi-agent outputs and does not contain any attempts to bypass core safety guardrails.
Audit Metadata