Show HN：TDAD - 開源的測試驅動開發工作流程，讓 AI 逐步完善程式碼

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

To see all available qualifiers, see our documentation.

Visual Test-Driven AI Development for AI Agents

License

zd8899/TDAD

Folders and files

Latest commit

History

Repository files navigation

TDAD: Test-Driven AI Development

TDAD (Test-Driven AI Development) is a visual workflow engine that upgrades your AI from a chaotic code generator into a disciplined engineer.

It enforces a Plan → Spec → Test → Fix cycle where runtime feedback (not just text) guides your AI to deliver working software, not just snippets.

Key Features: Local-first, Zero API calls, Use your own AI (Claude, ChatGPT, Cursor, etc.) and Free.

Quick Links: Installation • Getting Started • Features • Contributing • Community

The Problem with AI Coding (And How We Fix It)

Most AI coding tools fail because they lack runtime feedback and engineering discipline. TDAD fixes the 4 major flaws of AI development:

1. Problem: "Getting Lost in the Chat"

AI chats are linear and forgetful. As you build more features, you lose track of the big picture, what's actually finished, and how it all connects.

2. Problem: The "Lazy Prompt" Effect

Most AI code fails because the prompt was vague. You shouldn't have to write paragraphs of context to get a simple feature.

3. Problem: The "Code Snippet" Trap

You ask for a feature, the AI generates 5 files, and... nothing works. The imports are wrong, and the logic is hallucinated.

4. Problem: The Debugging Loop of Death

When code fails, you paste the error back to the AI. It guesses blindly, often breaking five other things in the process.

Privacy & Control

Installation

Prerequisites

Install from VS Code Marketplace

Manual Installation (Development)

Getting Started

Quick Start (New Project)

Quick Start (Existing Project)

Your First Feature

Features

1. The "Canvas" System (Visual Workflow Board)

Instead of generating code blindly, TDAD provides a visual canvas to plan and track your features.

2. The "Interactive Workflow" (The Core Experience)

TDAD does not call OpenAI/Claude for code generation. Instead, it serves as a Prompt Engineering Platform with a 4-step linear pipeline.

The Bottom Action Bar displays when a feature node is selected, showing a strict TDD workflow:

Step 1: BDD (Always available - the starting point)

Step 2: Tests (Enabled ONLY if BDD exists)

Step 3: Run (Enabled ONLY if Tests exist)

Step 4: Fix (Enabled ONLY if Tests exist)

Key Benefits:

3. The "Project Wizard" (Onboarding & Setup)

We solve the "Blank Page Problem" with two distinct workflows accessible from the Welcome Overlay.

Option A: Start New Project (3-step wizard)

Step 1: Define & Document

Step 2: Scaffold Structure

Step 3: Generate Blueprint

Option B: Map Existing Codebase (Reverse engineer mode)

4. The Dependency System (Reusing Actions)

When Feature B depends on Feature A, avoid duplicating logic. Instead, import and call the action function from the dependency.

5. The "Golden Packet" (Fixing Tests)

When a test fails, TDAD provides the "Golden Packet" to help the AI fix it.

The Golden Packet contains these sections:

SYSTEM RULES: FIX MODE - AI instructions emphasizing:

Scaffolded Files - Paths to read:

Project Context - Tech stack detected from package.json (React, Next.js, Playwright, etc.)

Dependencies - For each upstream feature:

Documentation Context - Optional files linked by user:

Previous Fix Attempts - What was already tried (automated mode only):

TEST RESULTS - The critical debugging data:

Your Task - Clear instructions: Read specs → Use trace → Fix APP → Verify

Checklist - Pre-flight checks before submitting fixes

6. The "Orchestrator" (Test Runner)

TDAD runs the loop.

7. "Auto-Pilot" (Lazy Mode)

Auto-Pilot (aka "Lazy Mode") automates the repetitive loop of BDD → Test → Fix by orchestrating your CLI agents (Claude, Cursor, etc).

Status: Available Now (Free)

How it works:

Note: Auto-Pilot works with your own CLI agents (Claude Code, Aider, etc.). TDAD doesn't accept or store any API keys - you use your preferred agent with your own credentials and security policies.

Technical Architecture

1. Smart Scaffolding (Filesystem)

2. TDAD Fixtures (Centralized Trace Capture)

3. The Prompt Library (The Protocol)

4. The Cortex (Dynamic Context Engine)

The Cortex is the brain that feeds the "Golden Packet".

Contributing

Contributions are welcome! TDAD is open source and community-driven.

How to Contribute

Areas for Contribution

Development Setup

See the Installation section for manual installation steps. After cloning:

License

TDAD is licensed under the MIT License. See the LICENSE file for full details.

Community

Join the discussion:

Support

Commercial Support

For enterprise support, training, or custom integrations, contact us at [email protected]

Built with ❤️ by the TDAD community

About

Visual Test-Driven AI Development for AI Agents

Topics

Resources

License

Uh oh!

There was an error while loading. Please reload this page.

Stars

Watchers

Forks

Releases

Packages

  0

Contributors

  3

Languages

Footer

Footer navigation