agent-assistant-design-team

Adversarial collaboration is a triad-driven design process where a tech lead, executor, and reviewer critique concepts to surface edge cases and enforce guardrails. This Golden Triangle approach reduces single-point failures and improves overall product design quality.

You orchestrate a multi-agent design team by loading the design-team workflow to coordinate agents across concept, architecture, and review stages. The system spawns and manages agents adaptively to validate iterative design decisions and apply constructive critique.

Multi-agent design orchestration requires Codex native multi-agent orchestration configured at ~/.codex/agents. It loads the workflow from ~/.codex/skills/agent-assistant/commands/design/team.md and resolves dependencies from ~/.codex/skills/agent-assistant/matrix-skills/.

Use triad-driven critique for product and software design tasks requiring iterative validation across concept, architecture, and review stages. It is ideal when you need to surface edge cases, establish guardrails, and reduce single-point failures in collaborative design scenarios.

The best way to coordinate multiple agents for design validation is using a Golden Triangle workflow with clear roles for tech lead, executor, and reviewer. This approach aligns agents to collaboratively design solutions while applying constructive critique to validate decisions.

Adversarial multi-agent design workflows are limited to environments supporting Codex native orchestration and require resolving skills from specific local directories. They are not suitable for simple design tasks that lack the complexity to benefit from triad-driven critique and iterative validation.