Skip to content

Which Path? Choose Your Workflow in 30 Seconds ​

Who is this for? Anyone about to start work with Bear Craft who wonders "which commands do I run, and in what order?". Answer two questions, follow one path.

Rule of thumb: one situation = one path = one short command sequence. You never need the other 100+ commands to get started.


1. The decision tree ​

                    What are you doing today?
                               │
        ┌──────────────┬───────┴────────┬──────────────────┐
        ▼              ▼                ▼                  ▼
  A new product    Existing code,    A bug to fix     A planned sprint
  from scratch     no backlog yet                     you want finished
        │              │                │             without babysitting
        ▼              ▼                ▼                  ▼
   GREENFIELD      BROWNFIELD        BUGFIX         AUTONOMOUS SPRINT
  /bear-workflow:init  /bear-project:reverse scan  /bear-qa:tdd   /bear-workflow:auto-sprint N
        │              │          (+ Quick Flow)          │
        ▼              ▼                ▼                  ▼
     Section 2      Section 3       Section 4          Section 5

  Lost at any moment?  →  /bear-workflow:status   (Section 6)

Not sure? Ask yourself:

  1. Is there already code? No → Greenfield. Yes → go to question 2.
  2. Is there already a backlog (epics and stories)? No → Brownfield. Yes → go to question 3.
  3. Is it a single bug? Yes → Bugfix. No → continue the sprint with /bear-workflow:start N, or hand it over to the Autonomous Sprint.

2. Greenfield: a new product ​

You have an idea and an empty repository. Follow the main path, one phase at a time:

/bear-workflow:init  →  /bear-workflow:plan  →  /bear-workflow:design  →  /bear-workflow:start 1
                →  /bear-workflow:implement  →  /bear-workflow:review  →  /bear-workflow:retro
StepCommandWhat you get
1/bear-workflow:initProject analysis, recommended track (Quick, Standard or Enterprise), and .bmad/sprint-status.yaml
2/bear-workflow:planPRD, personas and the generated backlog
3/bear-workflow:designTechnical specification and architecture
4/bear-workflow:start 1Sprint 1 prepared: stories selected and ready for development
5/bear-workflow:implementStories built with TDD (Red → Green → Refactor)
6/bear-workflow:reviewSprint review preparation
7/bear-workflow:retroRetrospective, then loop back to /bear-workflow:start 2

For a narrated, step-by-step version, read the Complete Workflow tutorial.


3. Brownfield: existing code, no backlog ​

You inherit (or already wrote) a codebase, but nobody documented what it does. Reverse-engineer the product first, then join the main path:

/bear-project:reverse scan  →  /bear-project:reverse prd  →  /bear-project:reverse stories
               →  /bear-project:reverse gaps  →  /bear-workflow:start
StepCommandWhat you get
1/bear-project:reverse scanInventory of the codebase: stack, structure, modules
2/bear-project:reverse prdA PRD rebuilt from what the code actually does
3/bear-project:reverse storiesUser stories inferred from existing features
4/bear-project:reverse gapsMissing features, missing tests, technical debt
5/bear-workflow:startFirst sprint planned from the gaps; continue with /bear-workflow:implement

See also the Existing Project setup guide.


4. Bugfix: reproduce first, then fix ​

A bug does not need a PRD. Use the lightest track and let the test drive the fix:

/bear-qa:tdd        ← write the test that reproduces the bug FIRST (it must fail)
/bear-workflow:init --quick    ← Quick Flow track: implement only
  1. Run /bear-qa:tdd and describe the bug. The test that reproduces it is written before any fix, and must fail.
  2. Run /bear-workflow:init --quick to force the Quick Flow track (no planning or design phases).
  3. Fix the code until the test passes, then refactor. The test stays in the suite as a regression guard: a fixed bug should never come back.

See the Bug Fixing guide for details.


5. Autonomous sprint: one command, start to merge ​

When the work is already planned, let Bear Craft drive the whole sprint:

/bear-workflow:auto-sprint N

Prerequisites

  • A backlog exists (from the Greenfield or Brownfield path).
  • Sprint N is planned.

What it does: it chains start, task breakdown, validation, implementation, pull request, CI, review, retrospective and finally merge. The run ends with the PR, a green CI and the merge. You review the result instead of typing each command.

Prefer control over speed? Stay on the manual main path of Section 2.


6. Where am I? ​

At any point, whatever the path:

/bear-workflow:status

It shows the current phase, the progress of the sprint and the recommended next action.

Single source of truth: .bmad/sprint-status.yaml holds the status of every story (backlog → ready-for-dev → in-progress → review → done, or blocked). It is created by /bear-workflow:init, and every command reads and updates it. If two views disagree, this file wins.


Cheat sheet ​

SituationStart withThen
New product/bear-workflow:init/bear-workflow:plan → /bear-workflow:design → /bear-workflow:start 1
Existing code, no backlog/bear-project:reverse scan/bear-project:reverse prd → /bear-project:reverse stories → /bear-project:reverse gaps → /bear-workflow:start
Bug/bear-qa:tdd/bear-workflow:init --quick
Planned sprint, hands-off/bear-workflow:auto-sprint NReview the PR
Lost/bear-workflow:statusFollow the recommended action