Guia Completo de Workflows

Como criar, executar e gerenciar workflows

8 min de leitura
guiaworkflowsreferência

AIOX Workflows Guide

Version: 1.0.0 Last Updated: 2026-02-02 Status: Active


Overview

AIOX Workflows are orchestrated sequences of agent activities that automate complex development processes. They provide structured, repeatable patterns for common development scenarios.

Key Concepts

| Concept | Description | |---------|-------------| | Workflow | A YAML definition that orchestrates multiple agents through a sequence of steps | | Phase | A logical grouping of related steps within a workflow | | Step | A single action performed by an agent within a workflow | | Transition | Movement from one step to the next, optionally with conditions | | State | Persistent tracking of workflow progress across sessions |


Workflow Types

By Project Type

| Type | Description | Use Case | |------|-------------|----------| | Greenfield | New projects from scratch | Starting a new application | | Brownfield | Existing projects | Enhancing or auditing existing code | | Generic | Any project type | Cross-cutting processes like story development |

By Scope

| Scope | Description | Examples | |-------|-------------|----------| | Fullstack | Complete application | greenfield-fullstack, brownfield-fullstack | | UI | Frontend only | greenfield-ui, brownfield-ui | | Service | Backend only | greenfield-service, brownfield-service | | Discovery | Analysis and audit | brownfield-discovery |


Available Workflows

Core Development Workflows

1. Story Development Cycle

ID: story-development-cycle Type: Generic Agents: SM → PO → Dev → QA

The most common workflow for iterative development:

┌─────────────────────────────────────────────────────────────┐
│                   Story Development Cycle                    │
│                                                              │
│  @sm: Create Story → @po: Validate → @dev: Implement → @qa  │
│         │                  │               │            │    │
│         ▼                  ▼               ▼            ▼    │
│     Draft Story       10 Checks       Code + Tests    Gate   │
└─────────────────────────────────────────────────────────────┘

Phases:

  1. Story Creation - SM creates the next story from backlog
  2. Story Validation - PO validates with 10-point checklist
  3. Implementation - Dev implements with tests
  4. QA Review - QA runs quality gate

When to use:

  • Any story development (greenfield or brownfield)
  • Complete cycle with validation and quality gate
  • When you need process traceability

2. Greenfield Fullstack

ID: greenfield-fullstack Type: Greenfield Agents: DevOps → Analyst → PM → UX → Architect → PO → SM → Dev → QA

Complete workflow for new full-stack applications:

Phases:

  1. Environment Bootstrap - DevOps sets up project infrastructure
  2. Discovery & Planning - Create project brief, PRD, specs, architecture
  3. Document Sharding - Break down docs for development
  4. Development Cycle - Iterative story implementation

When to use:

  • Building production-ready applications
  • Multiple team members involved
  • Complex feature requirements
  • Long-term maintenance expected

3. Brownfield Discovery

ID: brownfield-discovery Type: Brownfield Agents: Architect → Data Engineer → UX → QA → Analyst → PM

Complete technical debt assessment for existing projects:

Phases:

  1. Data Collection - System, database, frontend documentation
  2. Initial Consolidation - Draft assessment
  3. Specialist Validation - DB, UX, QA reviews
  4. Final Reports - Assessment + Executive report
  5. Planning - Epic and stories creation

When to use:

  • Migrating from Lovable/v0.dev
  • Complete codebase audit
  • Technical debt assessment before investment

Other Workflows

| Workflow | ID | Description | |----------|-------|-------------| | Greenfield UI | greenfield-ui | Frontend-only new projects | | Greenfield Service | greenfield-service | Backend-only new projects | | Brownfield Fullstack | brownfield-fullstack | Enhancing existing fullstack apps | | Brownfield UI | brownfield-ui | Enhancing existing frontends | | Brownfield Service | brownfield-service | Enhancing existing backends | | QA Loop | qa-loop | Quality assurance cycle | | Spec Pipeline | spec-pipeline | Specification refinement | | Design System Build | design-system-build-quality | Design system creation |


How to Create a Workflow

Step 1: Plan Your Workflow

Define:

  • Purpose: What problem does this workflow solve?
  • Agents: Which agents participate?
  • Sequence: What is the order of steps?
  • Conditions: Are there decision points or parallel activities?

Step 2: Use the Create Workflow Task

# Activate an agent that can create workflows
@architect

# Run the create workflow task
*create-workflow

Step 3: Answer Elicitation Questions

The task will ask:

  1. Target Context: core, squad, or hybrid
  2. Workflow Name: e.g., feature-development
  3. Primary Goal: What is the expected outcome?
  4. Stages/Phases: Main phases of the workflow
  5. Agent Orchestration: Which agents at each stage
  6. Resource Requirements: Templates, data files needed

Step 4: Workflow Structure

The generated workflow follows this structure:

workflow:
  id: my-workflow
  name: My Custom Workflow
  version: "1.0"
  description: "Description of what this workflow does"
  type: greenfield | brownfield | generic
  project_types:
    - web-app
    - saas

  metadata:
    elicit: true
    confirmation_required: true

  phases:
    - phase_1: Phase Name
    - phase_2: Another Phase

  sequence:
    - step: step_name
      id: unique-id
      phase: 1
      agent: agent-name
      action: Action description
      creates: output-file.md
      requires: previous-step-id
      optional: false
      notes: |
        Detailed instructions for this step...
      next: next-step-id

  flow_diagram: |
    ```mermaid
    graph TD
      A[Start] --> B[Step 1]
      B --> C[Step 2]
    ```

  decision_guidance:
    when_to_use:
      - Scenario 1
      - Scenario 2
    when_not_to_use:
      - Anti-pattern 1

  handoff_prompts:
    step1_complete: "Step 1 done. Next: @agent for step 2"

Step 5: Output Location

Workflows are saved based on context:

  • Core: .aiox-core/development/workflows/{name}.yaml
  • Squad: squads/{squad}/workflows/{name}.yaml
  • Hybrid: squads/{squad}/workflows/{name}.yaml

How to Run a Workflow

Method 1: Guided Mode (Default)

# Start a workflow
*run-workflow story-development-cycle start

# Check status
*run-workflow story-development-cycle status

# Continue to next step
*run-workflow story-development-cycle continue

# Skip optional step
*run-workflow story-development-cycle skip

# Abort workflow
*run-workflow story-development-cycle abort

Method 2: Engine Mode

# Run with full engine automation
*run-workflow greenfield-fullstack start --mode engine

Workflow State

State is persisted in .aiox/{instance-id}-state.yaml:

instance_id: "wf-abc123"
workflow_name: "story-development-cycle"
status: "active"
current_step: 2
total_steps: 4
steps:
  - id: create
    status: completed
    completed_at: "2026-02-02T10:00:00Z"
  - id: validate
    status: in_progress
  - id: implement
    status: pending
  - id: review
    status: pending

Multi-Session Continuity

Workflows persist across Claude Code sessions:

  1. User starts new session
  2. Activates @aiox-master
  3. Runs *run-workflow {name} continue
  4. System loads state, shows current step
  5. User executes step
  6. Returns and runs continue again

Workflow Patterns

AIOX detects common workflow patterns based on command history:

Detected Patterns

| Pattern | Trigger Commands | Agent Sequence | |---------|-----------------|----------------| | Story Development | validate-story-draft, develop, review-qa | PO → Dev → QA → DevOps | | Epic Creation | create-epic, create-story, validate-story-draft | PO → SM → Architect | | Architecture Review | analyze-impact, create-doc, review-proposal | Architect → QA → Dev | | Git Workflow | pre-push-quality-gate, github-pr-automation | Dev → DevOps | | Database Workflow | db-domain-modeling, db-schema-audit | Data Engineer → Dev → QA |

Pattern Detection

The system uses workflow-patterns.yaml to:

  • Detect which workflow you're in based on commands used
  • Suggest next steps with confidence scores
  • Provide contextual handoff messages

Best Practices

Workflow Design

  1. Keep phases focused - Each phase should have a clear purpose
  2. Define clear handoffs - Document what each agent passes to the next
  3. Include optional steps - Allow flexibility for simple cases
  4. Add decision guidance - Help users know when to use/not use

Workflow Execution

  1. Start with status - Check *run-workflow {name} status before continuing
  2. Follow handoff prompts - They contain important context
  3. Don't skip required steps - Only optional steps can be skipped
  4. Document decisions - Keep notes for future reference

Workflow Creation

  1. Test with simple cases first - Validate the flow works
  2. Include flow diagrams - Visual representation helps understanding
  3. Add detailed notes - Future users will thank you
  4. Define error handling - What happens when things go wrong?

Workflow vs Task

| Aspect | Workflow | Task | |--------|----------|------| | Scope | Multiple steps, multiple agents | Single step, single agent | | State | Persisted across sessions | Stateless | | Use Case | Complex processes | Atomic operations | | Location | .aiox-core/development/workflows/ | .aiox-core/development/tasks/ |


Troubleshooting

Common Issues

Workflow not found:

Error: Workflow '{name}' not found
  • Check the workflow name matches the file ID
  • Verify the target context (core/squad)

No active instance:

Error: No active workflow instance found
  • Start the workflow first with *run-workflow {name} start

Step not optional:

Error: Cannot skip non-optional step
  • Complete the step or abort the workflow

Getting Help

# List available workflows
ls .aiox-core/development/workflows/

# Validate a workflow
*validate-workflow {name}

# View workflow details
cat .aiox-core/development/workflows/{name}.yaml

Detailed Workflow Documentation

For complete documentation for each workflow, including detailed step-by-step guides, flowcharts, and implementation details, see:


Related Documentation


AIOX Workflows Guide v1.0 - Orchestrating AI-Human Collaboration

Falar com o Time