Skip to content

Codex Bmad Skills

BMAD skills and workflows for OpenAI Codex (App, CLI, Web): intent-based execution, YAML project state, and reusable skill packs for planning, architecture, sprint delivery,…

Install

$git clone https://github.com/xmm/codex-bmad-skills

Install command taken from the project's README or registry entry. Always review code before running it.

BMAD Skills for OpenAI Codex

BMAD Skills for OpenAI Codex adds a structured, intent-driven workflow to Codex for product discovery, planning, architecture, and implementation.

It installs reusable BMAD skills, introduces bmad:* intents, and keeps project workflow state in explicit YAML files inside your repository.

Original GitHub project: xmm/codex-bmad-skills

What This Project Does

This repository provides:

  • BMAD skills that can be installed into Codex
  • intent-based workflow routing via bmad:*
  • project-local workflow state in bmad/*.yaml
  • a repeatable path from idea to implementation

Use it when you want Codex to guide work through defined phases instead of relying on ad-hoc prompts.

Who This Is For

This project is for:

  • developers using OpenAI Codex as a primary coding assistant
  • teams that want a consistent AI-assisted delivery workflow
  • projects that need explicit planning and traceable workflow state

Quick Start

Requirements

  • yq v4+
  • python3 (or python)

The installers validate both dependencies before install.

Install

Install to the default global path

./installers/install-codex.sh

Install to a custom path (example: project-local)

./installers/install-codex.sh --dest "/.agents/skills"

Windows PowerShell

./installers/install-codex.ps1
./installers/install-codex.ps1 -Dest "\.agents\skills"

Notes:

  • --dest and -Dest are optional
  • the default destination is $HOME/.agents/skills
  • restart Codex after installation so it reloads newly installed skills

Start Codex

codex

Initialize BMAD in a Project

bmad:init name "My Project" communication language is English documentation language is English

Check Current Status

bmad:status
bmad:next

For a guided walkthrough, see Getting Started.

What Gets Created

After initialization, the project contains:

  • bmad/project.yaml - project settings and language preferences
  • bmad/workflow-status.yaml - current workflow phase and progress
  • bmad/sprint-status.yaml - sprint and delivery tracking
  • docs/bmad/ - generated BMAD documents
  • docs/stories/ - implementation stories

How The Workflow Works

BMAD in Codex is organized into four main phases.

1. Analysis

Use this phase to explore the problem space and define direction.

Typical intents:

  • bmad:product-brief
  • bmad:research
  • bmad:brainstorm

Typical outputs:

  • docs/bmad/product-brief.md
  • docs/bmad/research-report.md
  • docs/bmad/brainstorm.md

2. Planning

Use this phase to turn discovery into scoped plans and product documentation.

Typical intents:

  • bmad:prd
  • bmad:tech-spec
  • bmad:prioritize
  • bmad:ux-design

Typical outputs:

  • docs/bmad/prd.md
  • docs/bmad/tech-spec.md
  • docs/bmad/prioritization.md
  • docs/bmad/ux-design.md

Planning rules:

  • level 0-1 projects require tech-spec
  • level 2-4 projects require prd and architecture

3. Solutioning

Use this phase to define architecture and validate readiness before implementation.

Typical intents:

  • bmad:architecture
  • bmad:gate-check

Typical outputs:

  • docs/bmad/architecture.md
  • docs/bmad/gate-check.md

4. Implementation

Use this phase to plan delivery, create stories, build, and review.

Typical intents:

Similar plugs

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local…

Why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.

The 1st Collaborative AI Design Agent Workspace. Local-first and open source. Create prototypes, websites, apps, dashboards, presentations, images, and videos with your team…

Explore related