Getting Started

Copy AIDDbot in, then walk the delivery loop

AIDDbot is markdown for AI-Driven Development. One command copies it into your repo — no package dependency. It works with Cursor, Claude Code, GitHub Copilot, Codex, and other agent harnesses.

You invoke a public orchestrator. It composes the workflow and spawns Architect, Builder, or Craftsman where needed.

1. Copy AIDDbot into your project

From your project root (Node 18+):

npx --allow-git=all github:AIDDbot/AIDDbot init

That copies .agents/ and the adapters your editor needs. Existing files are left alone. Preview with --dry-run; replace differing files with --force.

Full install options live in the repo getting started guide.

2. Establish your solution

Run one entrypoint to set the foundation:

/architect-solution-foundation

On an existing app, Architect maps what is there. On an empty or docs-only repo, it designs first and scaffolds when you confirm. You approve the route when the choice is unclear.

3. Deliver a requirement

Describe what you need:

/build-requested-change riders can rate a trip 1 to 5 stars

Architect scopes the work into one or more specifications. You review problem, outcomes, and acceptance criteria when the workflow stops — that is your main checkpoint.

After approval, delivery plans, codes, verifies, qualifies, and ships. No extra slash command is required. Include YOLO in the prompt if you want to skip approval stops.

4. Improve what you already ship

Turn durable quality findings into a reviewed remediation:

/craft-lasting-quality

Craft gathers evidence, consolidates findings, and — after you accept the scope — fixes, verifies, qualifies, and ships a behavior-preserving patch. Include YOLO to skip the remediation checkpoint.

Whats next?

The usual loop after establishment:

  1. /build-requested-change — scope, specify, implement, verify, qualify, and ship
  2. /craft-lasting-quality — optional evidence-backed remediation

Next: Workflow · Skills catalog · GitHub