Skills Reference

The BMAD Planning & Orchestrator plugin ships 20 skills, not slash commands. Skills are namespaced under the plugin and invoked as:

/bmad-planning-orchestrator:<skill-name>

Most skills are auto-invoked by Claude based on what you are doing — you rarely type the full namespace. When Claude detects that your planning workflow calls for a particular skill, it will invoke it automatically. You can always call any skill explicitly if you prefer.

Attribution: The BMAD Method is created and maintained by the BMAD Code Organization. This plugin is an independent Claude Code harness — not an official BMAD product. All methodology credit belongs to the BMAD Code Organization.

What These Skills Do (and Do Not Do)

The plugin plans and orchestrates. It never writes application code, runs tests, lints, checks coverage, or reviews implemented diffs. The last artifact it produces is a ready-for-dev story file and a handoff-manifest.json for consumption by external dev tools.

Skills by Phase

Cross-Phase (Orchestration Spine)

Phase 1 — Analysis (Optional)

Phase 2 — Planning

Phase 3 — Solutioning

Phase 4 — Implementation Handoff

Cross-Phase & Meta

Tracks, Not Levels

Right-sizing uses tracks chosen interactively during bmad-init. There are no numbered Levels 0–4.

Tracks are not locked. Scope growth can promote a project from Quick Flow to BMad Method (by running bmad-prd and bmad-architecture), or from BMad Method to Enterprise (by adding security and DevOps planning passes before the readiness gate). Record the change in decision-log.md.

Sizing: One Dev-Day Decomposition

There are no Fibonacci story points, no velocity, and no burndown charts in this plugin. Stories are sized to one developer-day. Progress is tracked by count (stories delivered vs. stories remaining). Sprint sequencing assigns dependency-ordered parallel waves, not capacity-constrained sprints.

Skill Detail

bmad-help

The orchestration spine. Reads the artifacts present in your workspace and routes you to the next appropriate skill. Use this any time you are unsure what to do next.

Auto-invoked: Yes — Claude calls this at the start of planning sessions.

Output: Status display; no file written.

bmad-init

Initializes the planning workspace. Interactively selects a track (Quick Flow / BMad Method / Enterprise), creates config.yaml, decision-log.md, and project-context.md. Can also be called with an Update intent to promote a track.

Auto-invoked: Yes — on first planning interaction in a new workspace.

Output: bmad-output/config.yaml, bmad-output/decision-log.md, bmad-output/project-context.md

bmad-brainstorm

Structured ideation session using techniques such as SCAMPER, SWOT, 5 Whys, and Six Thinking Hats. Parallel subagents apply different frameworks simultaneously for diverse perspective coverage.

Output: bmad-output/brainstorming-report.md

bmad-research

Live, web-sourced research: market, competitive, domain, and technical. Produces a cited research report. Parallel subagents cover different research dimensions concurrently.

Output: bmad-output/research-report.md

bmad-product-brief

Guided discovery conversation (persona: Mary, Business Analyst). Produces a product brief covering problem space, target audience, competitive landscape, and business model. Supports Create / Update / Validate intents.

Output: bmad-output/product-brief.md

bmad-prfaq

Amazon Working-Backwards press release + internal and external FAQs. Stress-tests the concept before a PRD is written.

Output: bmad-output/prfaq.md

bmad-spec

Five-field distiller: Problem, Capabilities, Constraints, Non-Goals, Success Metrics. Accepts any messy input (emails, notes, conversations) and extracts a clean SPEC.md kernel that feeds the PRD or tech-spec.

Output: bmad-output/SPEC.md

bmad-document-project

Read-only brownfield codebase scan. Produces current-state documentation as the planning ground truth for an existing project. Does not write any application code.

Output: bmad-output/project-documentation.md

bmad-prd

PRD facilitator (persona: John, Product Manager). Produces functional requirements (FR-### series), non-functional requirements (NFR-### series), a MoSCoW priority table, RICE scoring, and an epic outline. Supports Create / Update / Validate intents. Parallel subagents generate PRD sections concurrently.

Output: bmad-output/prd.md, bmad-output/prd-addendum.md, updates to bmad-output/decision-log.md

Prerequisite: product-brief.md (recommended) or direct input.

bmad-tech-spec

Lightweight planning document for Quick Flow track (1–15 story scope). Combines problem statement, requirements, technical approach, and story list into a single document. Replaces PRD + architecture for small-scope work.

Output: bmad-output/tech-spec.md

bmad-ux

Two-document UX contract (persona: Sally, UX Designer). DESIGN.md covers the visual system (design tokens, component specs, WCAG 2.1 AA compliance). EXPERIENCE.md covers user journeys and interaction flows. Optional; only run if the project has a UI.

Output: bmad-output/DESIGN.md, bmad-output/EXPERIENCE.md

Prerequisite: prd.md

bmad-architecture

System architecture design (persona: Winston, Architect). Produces architecture.md with component boundaries, ADRs, an FR/NFR coverage matrix, and explicit API/data model/naming conventions. Architecture conventions are the primary mechanism for semantic conflict prevention in parallel development.

Output: bmad-output/architecture.md

Prerequisite: prd.md

bmad-epics-and-stories

Shards the PRD and architecture into an ordered epics.md and compiled, self-contained story context files ({epic}.{story}.{slug}.story.md, each approximately 8K tokens). Each story carries Dev Notes, Acceptance Criteria, a Testing strategy, Tasks/Subtasks, a Dependency Map, and an Owned File/Module Scope. The Owned File/Module Scope is the primary mechanism for preventing file and merge conflicts in parallel development. Acceptance Criteria, Dev Notes, and Testing sections are locked for dev tools.

This is the last planning artifact the plugin produces before handoff.

Output: bmad-output/epics.md, bmad-output/stories/*.story.md

Prerequisites: prd.md, architecture.md

bmad-readiness-check

Gate check before story compilation begins. Cross-references PRD requirements against architecture coverage and epic completeness. Returns a PASS / CONCERNS / FAIL verdict. CONCERNS is addressable; FAIL blocks progression until resolved.

Output: bmad-output/readiness-report-<project-slug>-<date>.md

Prerequisites: prd.md, architecture.md

bmad-sprint-planning

Sequences stories by dependency order and assigns them to parallel waves. Produces sprint-status.yaml. No story points, no velocity, no burndown — progress is tracked by story count.

Output: bmad-output/sprint-status.yaml

Prerequisites: epics.md, stories/

bmad-parallel-plan

Builds a dependency DAG and topologically sorts stories into conflict-free concurrent waves. Each wave specifies worktree branch names and merge order so a worktree-based dev runner can execute the wave in parallel without file conflicts.

Output: bmad-output/parallelization-plan.md

Prerequisites: sprint-status.yaml

bmad-handoff

Emits the final plugin artifact: handoff-manifest.json, a stable versioned schema listing each ready story, its owned scope, its wave/parallel-set, and its dependencies. Any external dev tool or autonomous runner can consume this file directly.

Output: bmad-output/handoff-manifest.json

bmad-correct-course

Mid-stream scope or architecture change handler. Applies minimum-blast-radius re-planning: determines which downstream artifacts are affected and re-runs only those phases. Routes to bmad-epics-and-stories, bmad-sprint-planning, or bmad-parallel-plan as needed. Never re-codes.

bmad-investigate

Forensic bug triage. Grades evidence (A/B/C confidence), ranks hypotheses, and produces an investigation case file. Optionally emits a fix story for dev handoff.

Output: bmad-output/investigation-*.md, optional stories/*.story.md

bmad-builder

Meta-skill for extending the plugin. Scaffolds and validates new planning and orchestration skills. Enforces the scope law — builder skills created here must stay within planning/orchestration and must not generate application code.

Migration Map: Old Slash Commands to Skills

If you used the previous slash-command interface, this table shows the equivalent plugin skill.

Typical Flows by Track

Quick Flow

bmad-init  →  bmad-spec (optional)  →  bmad-tech-spec
          →  bmad-epics-and-stories  →  bmad-handoff

BMad Method

bmad-init  →  bmad-product-brief (optional)  →  bmad-prd
          →  bmad-architecture  →  bmad-ux (if UI)
          →  bmad-readiness-check (gate)
          →  bmad-epics-and-stories  →  bmad-sprint-planning
          →  bmad-parallel-plan (optional)  →  bmad-handoff

Enterprise

Everything in BMad Method, plus security and DevOps planning addenda added after bmad-architecture and before bmad-readiness-check.

Next Steps

• See getting started for installation and first-run instructions
• See subagent patterns for how parallel planning agents are coordinated
• See examples for complete end-to-end planning walkthroughs

The BMAD Method is created and maintained by the BMAD Code Organization. This plugin is an independent Claude Code harness for the method’s planning and orchestration workflows. No endorsement is implied. All methodology credit belongs to the BMAD Code Organization.