ToolExperimental

opc

Name: opc
Rating: 3.0 (184 reviews)
Author: iamtouchskyer

by iamtouchskyer

Single-skill multi-agent orchestrator with role-based QA and adversarial checks

Shell

Updated Jul 20, 2026

184

Stars

Forks

View on GitHub

Summary

Implements an adaptive multi-role agent orchestrator inside a single Claude Code skill. Provides 11 built-in roles, 4 orchestration modes, and adversarial quality-control loops to route work, validate outputs, and escalate when needed. Designed to run multi-agent workflows entirely within Claude with Hierarchical Multi-Agent Pattern and role-based delegation and automated checks. It also leverages Role-Based Agent Pattern to ensure clear role separation and responsibilities, aiding Delegation practices across tasks.

Key Benefits

As agents delegate subtasks, it's easy for errors or bad actors to slip through without role-aware validation. OPC makes role separation and adversarial QA first-class, so you can observe how individual roles perform and where failures occur. That visibility is useful for building agent track records and running continuous A2A evaluation across workflows, supported by the Agent-to-Agent Protocol (A2A).

Target Use Cases

Teams prototyping multi-agent workflows on Anthropic/Claude who need role separation, automated QA, and rapid iteration. For routing and coordination considerations within these workflows, consider the Dynamic Task Routing Pattern.

Use Cases

Orchestrating specialized roles (planner, reviewer, critic) inside a single Claude skill
Running adversarial quality checks to detect incorrect or low-quality agent outputs
Prototyping role-based delegation patterns before broader multi-model deployments

Need implementation help?

Expert guidance available