What happens if the two AI models cannot agree?

If convergence isn't reached within three rounds, the skill presents a structured comparison of the remaining conflicts for the user to make the final decision.

How does the negotiation process work?

It uses a three-phase protocol: Claude makes a proposal, Codex (GPT-5.4) provides a critical review with scores, and then Claude defends or revises the plan until convergence is reached.

What metrics are used to evaluate technical proposals?

Proposals are scored from 0-10 on implementation cost, maintainability, performance, scalability, and risk.

Can I involve the AI in a discussion about my existing code?

Yes, you can provide file paths and project context, allowing the models to analyze your actual implementation while debating design improvements.

Codex Negotiation

Name: Codex Negotiation
Author: CAPHTECH

byCAPHTECH

0•

协作工具

Facilitates structured architectural debates and technical decision-making between Claude and Codex to reach optimal consensus.

The Codex Negotiation skill establishes a formal protocol for resolving complex technical trade-offs and design dilemmas by leveraging multi-model reasoning. By following a rigorous three-phase process—Proposal, Critique, and Revision—it ensures that architectural choices are vetted against critical metrics like implementation cost, maintainability, and scalability. This skill is particularly valuable for developers who need to challenge their own assumptions or require a structured 'second opinion' to compare multiple technical approaches with quantitative scoring and evidence-based counter-proposals.

主要功能

01Comparison reports highlighting consensus points and trade-offs

02Context-aware analysis of local project files and source code

030 GitHub stars

04Structured three-phase convergence workflow (Proposal, Critique, Revision)

05Formal debate protocol between Claude and GPT-5.4 (Codex)

06Quantitative scoring across five key technical evaluation metrics

使用场景

01Refining system architectures to identify hidden risks and bottlenecks

02Resolving team design disagreements by generating a neutral trade-off analysis

03Selecting between competing libraries or frameworks based on performance and cost

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add caphtech/claude-marketplace codex-negotiation

For use in Claude.ai and ChatGPT

主要功能

01Comparison reports highlighting consensus points and trade-offs

02Context-aware analysis of local project files and source code

030 GitHub stars

04Structured three-phase convergence workflow (Proposal, Critique, Revision)

05Formal debate protocol between Claude and GPT-5.4 (Codex)

06Quantitative scoring across five key technical evaluation metrics

使用场景

01Refining system architectures to identify hidden risks and bottlenecks

02Resolving team design disagreements by generating a neutral trade-off analysis

03Selecting between competing libraries or frameworks based on performance and cost

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add caphtech/claude-marketplace codex-negotiation

For use in Claude.ai and ChatGPT