workflows-expert

# Workflows MCP Expert Skill

Activate when requests involve workflow execution, multi-step task orchestration, or CI/CD automation using the workflows-mcp MCP server.

## Activation Triggers

Activate when user requests mention:

- "run workflow", "execute workflow", "list workflows"
- "orchestrate tasks", "multi-step process", "DAG execution"
- "CI/CD pipeline", "git automation", "automated testing"
- "workflow information", "workflow details"
- Task coordination with dependencies
- "Chain commands", "automate workflow"

Common user phrases: "run tests before deploying", "create build pipeline", "execute only if previous succeeds", "run tasks in parallel"

## Core Workflow Pattern

### Standard execution: Discover → Inspect → Execute

Always use tag-based discovery instead of guessing workflow names. Inspect workflows before execution to understand required inputs.

### Step 1: Discover by Tags

Use `list_workflows` with tags (AND logic - workflows must have ALL tags):

```text
Tool: list_workflows
Parameters:
  tags: ['python', 'ci']
  format: 'markdown'
```

**Common tag combinations:**

- Python: `['python']`, `['python', 'testing']`, `['python', 'ci']`
- Git: `['git']`, `['git', 'commit']`, `['git', 'checkout']`
- CI/CD: `['ci']`, `['ci', 'deployment']`
- TDD: `['tdd']`, `['tdd', 'phase1']`

### Step 2: Inspect Workflow

Call `get_workflow_info` before executing to understand inputs, outputs, and structure:

```text
Tool: get_workflow_info
Parameters:
  workflow: 'python-ci-pipeline'
  format: 'markdown'
```

### Step 3: Execute Workflow

```text
Tool: execute_workflow
Parameters:
  workflow: 'python-ci-pipeline'
  inputs: {project_path: '/path/to/project'}
  response_format: 'minimal'
```

**Response status values:**

- `success` - Workflow completed successfully
- `failure` - Workflow failed (check `error` field)
- `paused` - Workflow paused for user input (use `resume_workflow`)

## Available MCP Tools

### Discovery & Information

**list_workflows(tags, fo