collab-board

What problem does it solve?

Single AI models have blind spots and can confidently state wrong answers. This Skill creates a strict turn-based collaboration between a PRIMARY and a SECONDARY AI model, enforcing skeptical peer review and gated phases so mistakes are caught early before they compound into larger failures.

Core Features & Use Cases

• Strict Alternating Turns: A PRIMARY and SECONDARY model take turns through PLAN and IMPLEMENTATION phases, with each turn reading only the files it needs to keep context small regardless of session length.
• Gated Phases: The PLAN phase cannot advance to IMPLEMENTATION until both models agree and all open points are resolved, preventing wrong decisions from becoming the foundation for later work.
• Skeptical Review: Each model must back claims with evidence and challenge the other, reducing confidently-wrong answers and shared blind spots through adversarial collaboration.
• Use Case: Have Claude and Codex collaborate on a feature, bug fix, refactor, investigation, or design — from planning through implementation — with full audit trails and bounded per-turn context.

Quick Start

Use the collab-board skill to start a new collaboration session between two AI models for your task, then drive the strict turn-based planning and implementation loop until both phases agree and the work is complete.