Skill Workflow Builder

Skill Workflow Builder

This skill provides an interactive utility script and guidelines for creating structured, multi-step workflow skills.

About Workflows

Workflow skills are structured step-by-step procedural prompt instructions that guide agents through complex, multi-tool operations. Workflows reside in: universal_skills/workflows/{domain}/{workflow_name}/

Step Schema

Each step in a workflow skill must follow a strict topological definition:

### Step N: [depends_on: Step A, Step B]

Expected: 

Where:

• Step N: Unique 0-indexed integer (e.g. Step 0, Step 1).
• **`**: The MCP tool or agent capability utilized (e.g., portainer-mcp, github-tools, systems-manager, user-interaction`).
• depends_on (Optional): A list of steps that must complete before this step is executed.

---

Interactive Workflow Builder Process

When building a new workflow skill, follow these sequential phases interactively with the user:

Phase 1: Review & Discover

1. Index existing workflows to find similar structures using the CLI tool:

``bash python universal_skills/agent-tools/skill-workflow-builder/scripts/build_workflow.py list ``

1. Verify available MCP tools and databases using query-kg or inspecting configurations.

Phase 2: Topological Brainstorming

1. Design the step-by-step procedure interactively with the user.
2. Ask:

• What is the domain/category (e.g., infra, dev-workflows, ops, finance, health, social, system, research)?
• What components/tools are needed for each step?
• What are the dependencies?

1. Draw a Mermaid diagram of the workflow DAG for the user to visually inspect and approve the flow.

Phase 3: Scaffolding

1. Run the scaffolding script in interactive mode to register the steps, tags, and description:

``bash python universal_skills/agent-tools/skill-workflow-builder/scripts/build_workflow.py scaffold --domain --description "" --interactive ``

1. The script will automatically:

• Check the steps topological structure for circular dependencies.
• Create the directory at universal_skills/workflows/{domain}/{workflow_name}/.
• Write a compliant SKILL.md file.
• Generate a default swarm orchestration references/team.yaml.

Phase 4: Validation

1. Verify the workflow skill compiles perfectly using:

``bash python universal_skills/agent-tools/skill-workflow-builder/scripts/build_workflow.py validate universal_skills/workflows/{domain}/{workflow_name}/SKILL.md ``

1. Save the final files in the user's source repository!

---

Topological Reference

Workflows are executed by a Parallel Orchestration Engine that parses the steps into a Directed Acyclic Graph (DAG). It is critical that:

• Step numbers do not duplicate.
• Dependencies reference actual prior steps.
• No cycles exist (e.g. Step A depends on Step B, which depends on Step A). The validator script will automatically catch these issues.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Knuckles-Team
• Source: Knuckles-Team/universal-skills
• License: MIT

Install and usage instructions live in the source repository linked above.