Create Workflow Command

Features

• Orchestrate multi-step execution with sub-agents
• Utilize file-based task prompts for detailed step instructions
• Manage context isolation for sub-agents
• Generate structured workflow command definitions
• Promote modularity and reusability in agent workflows

Use Cases

• Structuring complex feature implementation pipelines
• Defining multi-phase research and development processes
• Automating sequential tasks with distinct agent roles
• Building predictable agent workflows with clear boundaries

Non-Goals

• Directly executing workflows; it generates the commands for them
• Containing the LLM's context window for the entire workflow
• Replacing the need for user-defined tasks and goals

Workflow

1. Gather requirements for the workflow (name, description, steps, mode, agent type).
2. Create directory structure for tasks and optional agents.
3. Create task files for each step with detailed instructions and expected output.
4. Create the orchestrator command file using provided patterns and frontmatter.
5. Optionally, create custom agent definitions if needed.
6. Update plugin manifest if necessary and test the command.

Practices

• Workflow design
• Agent orchestration
• Context management

Prerequisites

• Access to an AI agent environment supporting commands and markdown
• Ability to create and manage files within a plugin directory