What is the 'Explore-Before-Prompt' pattern?

It is a strategy where the orchestrator spawns an 'explorer' agent to gather context on unfamiliar files before spawning an 'implementer' agent, ensuring the implementation prompt is highly accurate.

How does the Orchestrate phase prevent errors?

In the Orchestrate phase, the agent is restricted from using Edit or Bash tools. This forces the orchestrator to focus on planning and spawning specialized subagents, ensuring a clean separation of concerns.

How are code reviews handled in the workflow?

Once implementation and testing are complete, the workflow moves to a Review phase where it triggers external reviews (like Greptile) and feeds issues back into the task queue for automated fixing.

What does the /iterate command do in Claude Code?

The /iterate command initiates an autonomous TDD (Test-Driven Development) loop. It manages the transition between writing tests, implementing features, and running verifications until your exit conditions are met.

Can this skill handle multiple tasks at once?

Yes, the orchestrator can spawn multiple subagents in parallel to process the task queue, allowing for simultaneous implementation of different modules or features.

Iterative TDD Workflow Orchestrator

Name: Iterative TDD Workflow Orchestrator
Author: c-daly

byc-daly

0•

生産性とワークフロー

Automates test-driven development loops and parallel agent orchestration to ensure code quality and autonomous task completion.

This skill implements a sophisticated Test-Driven Development (TDD) workflow for Claude Code using an orchestrator-worker architecture. It manages the entire development lifecycle through structured phases—including Intake, Orchestrate, Test Writing, Implementation, Testing, and Review—while enforcing strict tool restrictions to maintain architectural integrity. By managing a dynamic task queue and spawning autonomous subagents for parallel execution, it allows developers to handle complex refactors and feature implementations with built-in quality gates, coverage requirements, and automated review cycles.

主な機能

01Enforced tool restrictions to separate planning from implementation

02Explore-Before-Prompt pattern for context-aware subagent delegation

03Integrated verification for pytest, linting, and coverage standards

04Autonomous TDD loop with automated phase transitions and gates

05Multi-agent orchestration for high-concurrency parallel task execution

060 GitHub stars

ユースケース

01Managing autonomous agent swarms to resolve technical debt through a systematic task queue

02Refactoring legacy codebases while ensuring strict adherence to test coverage and linting rules

03Executing complex feature additions that require coordinated changes across multiple modules

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add c-daly/agent-swarm iterate

For use in Claude.ai and ChatGPT