These flows are starting points. Copy one into .orkestra/workflow.yaml, adapt the prompt templates to your project, and run ork trak create --flow <name> to use it.

For the full list of available fields, see the Workflow Reference.

Full product feature

Best for complex features: upfront planning, parallel implementation via Subtraks, automated checks, and a thorough review before merge. This is the structure the default flow from ork init follows.

After the Breakdown stage is approved, Orkestra spawns parallel Subtraks — one per item in the breakdown — using the subtask flow. All Subtraks complete before the parent Trak advances to Work.

The Compound stage is marked is_automated: true, so it runs without a human approval step.

version: 2

flows:
  default:
    stages:
      - name: planning
        artifact: plan
        description: Clarifies requirements and produces an implementation plan
        prompt: planner.md
        model: opus
        capabilities:
          ask_questions: true

      - name: breakdown
        artifact: breakdown
        description: Researches the codebase and breaks the plan into parallel Subtraks
        prompt: breakdown.md
        capabilities:
          subtasks:
            flow: subtask  # Simpler flow used for child Traks (Subtraks)

      - name: work
        artifact: summary
        description: Implements code changes in an isolated worktree
        prompt: worker.md
        gate:
          command: "$ORKESTRA_PROJECT_ROOT/.orkestra/scripts/checks.sh"
          timeout_seconds: 600

      - name: review
        artifact: verdict
        description: Reviews completed work and produces an approve/reject verdict
        prompt: reviewer.md
        model: opus
        capabilities:
          approval:
            rejection_stage: work

      - name: compound
        artifact: learnings
        description: Captures learnings to improve future Traks
        prompt: compound.md
        is_automated: true

    integration:
      on_failure: work

  subtask:
    stages:
      - name: work
        artifact: summary
        prompt: worker.md
        gate:
          command: "$ORKESTRA_PROJECT_ROOT/.orkestra/scripts/checks.sh"
          timeout_seconds: 600

      - name: review
        artifact: verdict
        prompt: reviewer.md
        capabilities:
          approval:
            rejection_stage: work

    integration:
      on_failure: work

The prompt templates used here — planner.md, breakdown.md, worker.md, reviewer.md, and compound.md — are created by ork init. The subtask flow reuses worker.md and reviewer.md from the default flow.

Quick fix

Best for bug fixes, small changes, documentation typos, and dependency updates. No planning or breakdown — just implement and review.

# In .orkestra/workflow.yaml, add under flows:
flows:
  quick:
    stages:
      - name: work
        artifact: summary
        description: Implements the change and produces a summary
        prompt: worker.md
        gate:
          command: "$ORKESTRA_PROJECT_ROOT/.orkestra/scripts/checks.sh"
          timeout_seconds: 300

      - name: review
        artifact: verdict
        description: Reviews the fix and produces an approve/reject verdict
        prompt: reviewer.md
        capabilities:
          approval:
            rejection_stage: work

    integration:
      on_failure: work

This flow reuses worker.md and reviewer.md — no additional prompt templates needed.

To create a Trak using this flow:

ork trak create \
  -t "Fix null pointer in auth middleware" \
  -d "Line 47 of src/auth/middleware.rs panics when the session token is missing." \
  --flow quick

Documentation writer

Best for writing or updating documentation, READMEs, API references, and migration guides. No gate script — documentation doesn’t compile.

# In .orkestra/workflow.yaml, add under flows:
flows:
  docs:
    stages:
      - name: research
        artifact: findings
        description: Explores the codebase and clarifies what needs documenting
        prompt: docs-researcher.md
        capabilities:
          ask_questions: true

      - name: draft
        artifact: draft
        description: Writes the documentation based on research findings
        prompt: docs-writer.md

      - name: review
        artifact: verdict
        description: Reviews the draft for accuracy, completeness, and clarity
        prompt: docs-reviewer.md
        capabilities:
          approval:
            rejection_stage: draft

    integration:
      on_failure: draft

This flow uses custom prompt templates that ork init does not create. Add these files to .orkestra/agents/:

• docs-researcher.md — instruct the agent to explore relevant source files, understand the feature being documented, and identify what the documentation must cover. Include questions about audience, depth, and scope.
• docs-writer.md — instruct the agent to write complete, accurate documentation based on the research artifact. Specify the format (MDX, Markdown, etc.) and any style guide requirements.
• docs-reviewer.md — instruct the agent to evaluate the draft against the research findings for accuracy, check for missing sections, and produce an approve/reject verdict with specific feedback.

See Prompt Templates for guidance on writing effective stage prompts.

Storybook story generator

Best for generating Storybook stories for existing components. A custom gate script verifies that stories build correctly before human review.

# In .orkestra/workflow.yaml, add under flows:
flows:
  storybook:
    stages:
      - name: analyze
        artifact: analysis
        description: Identifies which components need stories and their required variants
        prompt: storybook-analyzer.md
        capabilities:
          ask_questions: true

      - name: generate
        artifact: summary
        description: Generates Storybook stories for each identified component
        prompt: storybook-writer.md
        gate:
          command: "$ORKESTRA_PROJECT_ROOT/.orkestra/scripts/storybook-check.sh"
          timeout_seconds: 120

      - name: review
        artifact: verdict
        description: Reviews stories for accuracy, completeness, and visual correctness
        prompt: reviewer.md
        capabilities:
          approval:
            rejection_stage: generate

    integration:
      on_failure: generate

Create .orkestra/scripts/storybook-check.sh:

#!/bin/bash
set -e
cd "$ORKESTRA_WORKTREE_PATH"
pnpm storybook build --quiet

This flow uses two custom prompt templates:

• storybook-analyzer.md — instruct the agent to explore the component library, identify components lacking story coverage, and produce a structured list of required variants (default, loading, error, edge cases).
• storybook-writer.md — instruct the agent to write complete CSF3 stories based on the analysis artifact, matching the project’s existing story patterns.

The review stage reuses reviewer.md from the default flow.

Using multiple flows

All flows live in a single .orkestra/workflow.yaml under the flows: map. Add each flow as a named entry:

version: 2

flows:
  default:
    # ...
  quick:
    # ...
  docs:
    # ...
  storybook:
    # ...

You can also share gate scripts across flows. The storybook-check.sh above is a separate file from checks.sh — reference it by path in the gate command field. See Shell Scripts for the gate script interface and available environment variables.