Skip to main content

Creating a Workflow

This guide walks you through building your first workflow in Revelry Ai.

Prerequisites​

Before creating a workflow, you need to have skills ready. Each workflow step maps to a skill.

Why skills first?

  • Each workflow step requires a specific prompt
  • Skills define the inputs and outputs for each step
  • Testing skills individually ensures quality before assembly

Step 1: Create Skills​

Create the skills you'll use for each step:

  1. Navigate to Tools > Skills
  2. Click Create Skill
  3. Define the prompt for each workflow step:
    • Step 1 Example: "Gather requirements from documents"
    • Step 2 Example: "Generate PRD draft based on requirements"
    • Step 3 Example: "Quality check and finalize PRD"
  4. Define inputs and parameters for each skill
  5. Test skills individually before building your workflow

Tip: Start with 2-3 skills for a simple workflow. You can add more complexity later.

Step 2: Navigate to Workflows​

Click Tools in the left navigation, then the Workflows tab.

Click Create Workflow to open the workflow builder.

Step 3: Build Your Workflow​

Use the workflow builder to configure your workflow:

Add Steps in Sequence​

  1. Click Add Step
  2. Select a skill for this step
  3. Map inputs from previous steps or workflow inputs
  4. Configure any step-specific parameters
  5. Repeat for each step in your workflow

Map Inputs and Parameters​

Define how data flows through your workflow:

  • Workflow inputs - Variables provided when triggering the workflow
  • Step inputs - Data from previous steps or workflow inputs
  • Step outputs - Results that can be used by subsequent steps

Example workflow:

Step 1: Research (Input: topic) → Output: research_data
Step 2: Generate (Input: research_data) → Output: draft_content
Step 3: Refine (Input: draft_content) → Output: final_artifact

Configure Data Flow​

Specify how information moves between steps:

  • Use step outputs as inputs for subsequent steps
  • Pass workflow-level inputs to any step
  • Chain steps together logically

Step 4: Save Your Workflow​

Before saving, review your configuration:

  1. Review all steps - Verify the sequence makes sense
  2. Check input mapping - Ensure data flows correctly
  3. Provide a clear name - Describe what the workflow does
    • Good: "PRD Generator with QA"
    • Less helpful: "Workflow 1"
  4. Associate with project(s) - Select relevant projects
  5. Click Save Workflow

Step 5: Test Your Workflow​

Before using your workflow for production:

  1. Trigger a test execution
    • Provide sample inputs
    • Watch the workflow execute step-by-step
  2. Review generated artifacts
    • Check output quality
    • Verify all steps completed successfully
  3. Refine as needed
    • Adjust skills
    • Modify input mappings
    • Add or remove steps
  4. Test again until you're satisfied

Example: Simple PRD Generator​

Workflow goal: Generate a Product Requirements Document with built-in quality checks

Skills needed:

  1. "Requirements Gatherer" - Extract requirements from documents
  2. "PRD Generator" - Create PRD from requirements
  3. "Quality Checker" - Validate PRD completeness

Workflow configuration:

Name: "PRD Generator with QA"
Project: "Product Management"

Step 1: Requirements Gatherer
  - Input: product_name (workflow input)
  - Input: feature_description (workflow input)
  - MCP Tools: Enabled (query documents)
  - Output: requirements_list

Step 2: PRD Generator
  - Input: requirements_list (from Step 1)
  - Input: product_name (workflow input)
  - Output: prd_draft

Step 3: Quality Checker
  - Input: prd_draft (from Step 2)
  - Input: requirements_list (from Step 1)
  - Output: final_prd (artifact)

Best Practices​

Start Simple​

Begin with 2-3 steps before adding complexity. Get the basic workflow working, then enhance.

Use Clear Naming​

Name each step descriptively:

  • Good: "Gather Requirements", "Generate Draft", "QA Check"
  • Less helpful: "Step 1", "Step 2", "Step 3"

Build in Validation​

Add quality check steps between major stages:

  • Verify data gathered before generation
  • Check draft quality before finalization
  • Validate outputs meet requirements

Test Skills First​

Always test individual skills before adding them to workflows. This ensures each step works correctly in isolation.

Common Patterns​

Research → Generate → Refine​

  1. Gather information from multiple sources
  2. Generate content based on research
  3. Refine and validate the output

Gather → Analyze → Report​

  1. Collect data from documents and tools
  2. Analyze patterns and insights
  3. Generate formatted report

Template → Populate → Validate​

  1. Load template structure
  2. Fill template with dynamic content
  3. Verify all sections are complete

Next Steps​