Ralph-Workflow is an open-source agent workflow harness and skill orchestration framework designed to standardize and automate software development cycles across AI coding agents. Maintained by the Ralph Workflow organization, the tool structures agent behavior using modular skill bundles, development plans, and structured artifact submission pipelines. Operating primarily through CLI tooling and specialized agent skill directories (supporting configurations across environments like Claude Code and OpenAI Codex), Ralph-Workflow manages end-to-end task flows. Day-to-day use involves assigning tasks to agents that follow explicit protocols for brainstorming, test-driven development (TDD), plan execution, code review requests, and Git worktree isolation. What sets Ralph-Workflow apart from standard interactive agent harnesses is its heavy emphasis on subagent-driven development, parallel agent dispatching, and rigid verification loops. Rather than treating an agent as a simple chat interface, it enforces repeatable software engineering rituals including structured commit cleanup, baseline UI checks, security reviews, and artifact verification. Ralph-Workflow is an open-source project hosted on GitHub, utilizing local workspace configurations and standardized `.agents/skills` trees to manage agent permissions and workflows.
Tags: agent harness, ai agent, cli, coding agent, developer tools, OpenSource
.agents, .claude, and .codex environments. - Subagent-driven development — Supports dispatching parallel subagents to execute discrete subtasks within isolated development tracks. - Built-in verification loops — Enforces verification steps, automated testing, and security review cycles prior to submitting development results. - Git worktree management — Integrates Git worktree isolation (using-git-worktrees) to run agent executions without polluting the main working directory. - Artifact submission pipelines — Standardizes execution outputs through structured artifacts for development results, plans, and commit messages. - Structured planning and TDD protocols — Features dedicated workflows for writing plans, step-by-step plan execution, and test-driven development.requesting-code-review and security-review skills. The agent inspects modified code against configured coding standards, checks for vulnerabilities, and outputs commit cleanup artifacts.Ralph-Workflow is built for software engineers, DevOps teams, and engineering organizations looking to enforce disciplined, repeatable agent workflows across their codebases. It is especially suited for developers who require structured planning, subagent orchestration, and strict verification loops rather than unconstrained single-prompt generation. It is less suitable for casual developers seeking an ad-hoc, prompt-and-chat coding assistant.
bash git clone https://github.com/Ralph-Workflow/Ralph-Workflow.git cd Ralph-Workflow (Note: Check the repository for specific package manager distribution or CLI binary setups as updates are released.)bash # Verify skills are in place within your project ls -la .agents/skills Configure local workspace permissions and ensure your model provider credentials (e.g., Anthropic or OpenAI API keys) are exported in your terminal environment.bash # Example invocation utilizing the test-driven development and planning skills ralph-workflow run --skill test-driven-development --plan "Implement user authentication middleware with JWT verification"Part of CLI Coding Agents
Categories: AI Coding Agents, Code Assistants, Developer Tools