What is a 'discriminating test' in this skill?

A discriminating test is a specific test case where one proposed approach passes while another fails, providing clear, empirical evidence for one solution over the other.

What happens if the tests don't produce a clear winner?

If tests do not discriminate between options, the skill preserves the dissenting views and presents the specific trade-offs to the user for a final human decision.

Which testing frameworks does Resolve support?

The skill is designed to work with common command-line testing tools, including native support for pytest and npm test suites.

How does Resolve prevent bias during the decision process?

It utilizes an orchestrator-subagent pattern to isolate contexts, ensuring that the agents designing or executing tests are not influenced by the rhetoric used in the initial proposals.

When should I use the /resolve command?

Use /resolve whenever agents disagree, when you are faced with multiple valid technical approaches, or when making high-stakes architectural decisions with significant trade-offs.

Resolve Adjudicator

Name: Resolve Adjudicator
Author: Mburdo

byMburdo

•

生産性とワークフロー

Resolves technical disagreements and architectural conflicts using evidence-based testing and multi-agent adjudication.

Resolve is a specialized Claude Code skill designed to settle disagreements between AI agents or competing technical approaches by prioritizing empirical evidence over rhetoric. Utilizing an orchestrator-subagent pattern, it systematically gathers positions, generates discriminating tests, executes them, and adjudicates the winner based on objective results. This skill is essential for maintaining architectural integrity and making data-driven decisions when multiple valid paths exist, ensuring that development remains unbiased and grounded in functional performance.

主な機能

01Orchestrator-subagent architecture to prevent context anchoring and bias during decision making.

0246 GitHub stars

03Systematic four-phase workflow: Position Gathering, Test Generation, Test Execution, and Adjudication.

04Automated generation of discriminating tests that highlight functional differences between approaches.

05Seamless integration with existing testing suites like pytest and npm test.

06Preservation of dissent and value trade-offs for human review when tests are non-discriminatory.

ユースケース

01Adjudicating between several valid refactoring paths based on performance and reliability metrics.

02Resolving conflicts when multiple AI agents propose different code implementations.

03Settling architectural disputes such as choosing between JWT or session-based authentication.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mburdo/knowledge_and_vibes resolve

For use in Claude.ai and ChatGPT

Download Skill