Skip to main content

Workflow Usage

This guide walks through using the core-agents agent team in a project after you have run ag-skill add core-agents.

Prerequisites​

  • ag-skill installed
  • core-agents bundle installed in your repo (ag-skill add core-agents)
  • An AI agent tool open (e.g. GitHub Copilot Chat, Cursor, Claude)

The spec-to-ship sequence​

Step 1 — Describe the feature to ag-architect​

Select the ag-architect agent and describe what you want to build. The architect will produce:

  • A structured specification document
  • A context map (docs/<feature>/context-map.md)
  • Architecture Decision Records under docs/<feature>/adr/
  • A summary of generated artifacts with file paths and a handoff prompt

Step 2 — Review and approve the spec​

Read through the spec. Ask the architect to revise anything unclear before proceeding to implementation.

In v1.0.0, the architect template explicitly waits for confirmation before handoff. Reply with approved (or equivalent, such as "looks good" or "proceed") when you are ready.

Step 3 — Hand off to ag-coder​

Switch to the ag-coder agent and reference the approved spec. The coder implements backend code following Clean Architecture and DDD conventions matching your chosen stack.

Step 4 — Hand off to ag-ui-developer​

Switch to the ag-ui-developer agent and reference the same spec. The UI developer implements the frontend components.

Step 5 — Run ag-tester​

Switch to the ag-tester agent. The tester writes unit, integration, and end-to-end tests for the implementation and enforces the TDD red-green-refactor cycle.

Tips​

  • Run all agents from the root of the repository.
  • Use ag-skill doctor to confirm skills are correctly installed before starting.
  • The workflow.md file installed in your agent directory contains the full sequence for reference.