dev_agent

Launch a Pantheon branch and run a TDD loop with builder and critic agents.

1|Updated Oct 29, 2025
One-click install
npx skills add https://github.com/IANTHEREAL/dev_agent --skill dev-agent
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: dev_agent
Source: https://github.com/IANTHEREAL/dev_agent/tree/main/.
Command: npx skills add https://github.com/IANTHEREAL/dev_agent --skill dev-agent

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill automates the orchestration of a multi-agent software engineering workflow using Pantheon MCP. It coordinates a builder (Codex or Claude Code) and a critic (review_code) to perform an auditable Implement → Review → Fix loop, generating branch lineages and work artifacts to reduce manual coordination, errors, and time to value.

Core Features & Use Cases

  • TDD-style loop orchestration: Each turn forks a new Pantheon branch; the chain remains linear and auditable.
  • Role-based automation: Builder and critic agents alternate to implement and verify code changes.
  • Artifact-driven traceability: Generates worklog.md and code_review.log capturing design decisions, tests, and reviewer findings for audits.

Quick Start

Run the dev_agent skill at the repository root to initiate a sample Implement → Review cycle, observe created branches, and inspect worklog.md and code_review.log in your workspace.

Frequently Asked Questions about dev_agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate a test-driven development loop with code review?

dev_agent automates a TDD loop by orchestrating builder and critic agents through Pantheon MCP. It forks a new branch per cycle, implements code changes, runs reviews, and fixes issues in a linear, auditable sequence without manual coordination between agents.

Can I use dev_agent to coordinate multiple agents on code changes?

Yes. dev_agent launches specialist agents in parallel to manage Implement → Review → Fix workflows. It routes work between builder agents (Codex or Claude Code) and a critic (review_code), maintaining strict linear branch lineage and returning both branch IDs and review artifacts.

What artifacts does dev_agent generate for audit and traceability?

dev_agent creates worklog.md capturing design decisions and tests, plus code_review.log recording reviewer findings. Both artifacts are exposed as outputs and embedded in downstream prompts, providing full visibility into the decision chain.

How does dev_agent integrate with Pantheon branching?

dev_agent polls Pantheon MCP to create branches rooted at a parent_branch_id and monitors until success or failure. It maintains linear branch lineage so each cycle forks cleanly, enabling repeatable, traceable code workflows without tangled history.

Do I need prior setup to run dev_agent in my repository?

Run dev_agent at the repository root to start. It requires access to Pantheon MCP and builder/critic agents but has no external dependencies. A quick run generates sample branches and logs you can inspect to verify the workflow.

What's the difference between dev_agent and manual code review workflows?

dev_agent eliminates manual agent coordination by automating role handoffs between builder and critic, forking branches per cycle, and capturing all decisions in worklog and review logs. This reduces errors, coordination overhead, and time to completion versus sequential manual steps.