Skip to content
Agentic Coding Toolkit

Quickstart

Learn how to configure ACT, choose the right workflow path, create a requirements document, break it into implementation tasks, and run the first task.

Make sure you’ve completed the Installation steps first.

1. Configure ACT in your project

Section titled “1. Configure ACT in your project”

Navigate to your project and start your supported AI coding client from that directory:

Terminal window
cd my_flutter_app
# Run one of these
claude
opencode
codex
copilot
agy
cursor .

Run act-config to choose where ACT saves specs and implementation tasks for your project:

Terminal window
/act-config

For projects without a GitHub origin, ACT defaults to local Markdown files. For projects with a GitHub origin, act-config offers GitHub Issues as the default place to save specs and implementation tasks. See Project Setup for the files act-config creates and the supported storage options.

2. Choose the workflow path

Section titled “2. Choose the workflow path”

Use plain skills for general work. Use wrapper -flutter skills when the project or task is Flutter/Dart.

General workFlutter/Dart work
/act-interview/act-interview-flutter
/act-create-spec/act-create-spec-flutter
/act-create-issues/act-create-issues-flutter
/act-implement/act-implement-flutter

3. Interview and create a Spec

Section titled “3. Interview and create a Spec”

A Spec is a requirements document that describes what should be built before code changes start.

Start by describing the feature:

Terminal window
/act-interview "create a shell script for distributing the iOS app to App Store Connect"

ACT will interview you to gather the requirements, constraints, and decisions that materially affect the work. When the interview is complete, create the Spec:

Terminal window
/act-create-spec

ACT will ask focused questions only where the answers materially affect scope, requirements, or implementation choices. ACT will save the Spec and Interview Ledger using your project setup.

Review the Spec. This is your chance to catch problems before any code is written.

4. Create Work Items

Section titled “4. Create Work Items”

Work Items are implementation tasks created from a Spec, with context, acceptance criteria, dependencies, and verification steps.

The example paths below use the default local Markdown setup. If your project uses a different local path or GitHub storage, use the Spec location from your project setup.

Terminal window
/act-create-issues ai_specs/0001-settings/spec.md

ACT analyzes your codebase for conventions and patterns, then creates Work Items with:

  • Bounded scope and acceptance criteria
  • Dependency information such as Blocked by
  • TDD test-first tasks where applicable
  • Verification expectations for each Work Item

Work Items are saved in the same storage location as their parent Spec. See Project Setup for local file paths and GitHub storage behavior.

5. Implement one Work Item

Section titled “5. Implement one Work Item”

Execute one Work Item:

Terminal window
/act-implement ai_specs/0001-settings/work-items/01-settings-screen.md

ACT will implement the Work Item, use TDD at agreed seams when feasible, run static analysis and focused tests during work, run the full suite before finishing, mark local acceptance criteria when complete, and commit intended files by default. Pass --do-not-commit to keep changes local.

What just happened?

Section titled “What just happened?”

You completed the full ACT workflow:

  1. Config — selected where ACT saves Specs and Work Items
  2. Interview — resolved important questions before writing requirements
  3. Spec — defined what to build with clarity
  4. Work Items — broke the Spec into independently executable units
  5. Implementation — executed one Work Item with verification

The key pattern is to resolve ambiguity before implementation, then give the agent a small task with explicit acceptance criteria and verification steps.

Next steps

Section titled “Next steps”