agent-assistant-test-team

Adversarial multi-agent testing activates a team of agents to critique each other and uncover workflow gaps. It evaluates team-based reasoning, rule adherence, and cross-agent memory interactions to stress-test AI decision pathways.

To start a multi-agent test session, invoke the /test:team workflow. This boots a three-agent Golden Triangle collaboration session that orchestrates adversarial testing to validate assistant responses and safeguards.

Yes, orchestrating adversarial agent testing relies on Codex runtime boot rules and Codex native agents. The testing framework also resolves skills from matrix-skills to properly govern the multi-agent test sequence.

A Golden Triangle team tests compliance by spawning three Codex-native agents to conduct adversarial collaboration. This evaluates how the Agent Assistant handles competing viewpoints while maintaining operational rule adherence.

The limitation of a three-agent team for workflow testing is its reliance on specific multi-agent orchestration environments. It requires Codex native agents and matrix-skills resolution, restricting use to compatible runtime scenarios.