What is the difference between EDD and TDD?

While similar to Test-Driven Development, EDD focuses on structured evaluation checks (Capability, Behavior, Non-Functional) that are specifically designed for AI agents to verify implementation against a standardized template.

Do I need a specific LLM to use the EDD skill?

No, the skill is model-agnostic and will utilize whatever model is configured in your global Claude settings or environment variables.

Where are the evaluation results and reports stored?

Reports are automatically saved to docs/actions/edd/ as Markdown and to .claude/metadata/actions/edd/ as JSON for programmatic access.

How does Swarm Mode work in the EDD skill?

Swarm Mode automatically spawns a team of specialized sub-agents (Coordinator and Teammates) to handle different check types in parallel, significantly speeding up the verification process.

EDD Framework (Eval-Driven Development)

Name: EDD Framework (Eval-Driven Development)
Author: alfredolopez80

byalfredolopez80

•

安全与测试

Implements a quality-first development workflow by defining structured evaluations before implementation and verifying results using multi-agent swarms.

The EDD (Eval-Driven Development) Framework is a specialized skill for Claude Code that enforces a rigorous 'define-before-implement' pattern. By providing structured evaluation templates and a multi-agent swarm mode, it allows developers to categorize requirements into Capability (CC), Behavior (BC), and Non-Functional (NFC) checks. This skill automates the verification process, ensuring that new features meet all predefined criteria before completion, while integrating seamlessly with the Ralph orchestration system for comprehensive project lifecycle management and automated reporting.

主要功能

01Model-Agnostic Core: Operates with any configured LLM including Claude, GLM-5, and Minimax.

02Multi-Agent Swarm Mode: Spawns specialized sub-agents to parallelize evaluation tasks for faster verification.

03Integrated Workflow: Seamlessly bridges the gap between specification, implementation, and verification phases.

04Structured Evaluation Framework: Uses CC, BC, and NFC prefixes to categorize and track requirements.

0599 GitHub stars

06Automated Action Reporting: Generates detailed markdown and JSON reports for every execution for full traceability.

使用场景

01Enforcing Test-Driven Development (TDD) principles through structured AI-ready evaluation specifications.

02Parallelizing complex feature verification across multiple specialized AI agents to reduce development bottlenecks.

03Maintaining high-quality documentation and audit trails for compliance using automatic action reporting.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add alfredolopez80/multi-agent-ralph-loop edd

For use in Claude.ai and ChatGPT

Download Skill