基于编程智能体的规格驱动开发
Spec-Driven Development with Coding Agents
Summary
Course landing page introducing the Spec-Driven Development short course by Paul Everitt at JetBrains in partnership with DeepLearning.AI. Covers what you’ll learn including writing project constitutions, feature specs, and applying the workflow to both new and legacy codebases.
Key Takeaways
- Spec-Driven Development (SDD) is the disciplined alternative to vibe coding, using clear markdown specs to guide coding agents
- Three core benefits: control large code changes with small spec changes, eliminate context decay, and improve intent fidelity
- The workflow: write a project constitution (mission, tech stack, roadmap), then iterate through plan-implement-validate feature loops
Course Overview
Welcome to Spec-Driven Development with Coding Agents, a short course built in partnership with JetBrains and DeepLearning.AI.
Instructor
Paul Everitt — Developer Advocate at JetBrains. Paul has extensive experience in developer tooling, Python web frameworks, and AI-assisted development workflows.
What You’ll Learn
This course teaches you how to bring engineering rigor to AI-assisted coding by writing clear specifications that guide coding agents. You will learn:
- How to write a project constitution that defines your mission, tech stack, and roadmap
- How to author feature specs that isolate scope and enable plan-implement-validate loops
- How to apply SDD to greenfield projects (starting from scratch)
- How to apply SDD to brownfield projects (existing legacy codebases)
- How to write your own agent skills to automate the spec-driven workflow
Why Spec-Driven Development?
Spec-Driven Development is currently the best type of workflow for building serious applications with agentic coding assistance. Rather than writing code by hand, you focus on writing down the context that the agent doesn’t already have — giving your coding agent a markdown file or a long prompt explaining exactly what to build.
Course Outline
- Introduction — Overview of SDD benefits and workflow
- Why spec-driven development? — Vibe coding vs. SDD, engineering rigor with agents
- Constitutions — Writing project-level constitutions for greenfield and brownfield projects
- Feature Specs — Authoring feature-level specifications with plan-implement-validate loops
- SDD in Practice — Hands-on application to real projects
- Agent Skills — Automating your SDD workflow with custom agent skills
基于编程智能体的规格驱动开发
摘要
课程介绍页，由 JetBrains 开发者布道师 Paul Everitt 主讲，与 DeepLearning.AI 联合推出。课程涵盖编写项目章程、功能规格说明，以及将 SDD 工作流应用于全新项目和传统遗留代码库。
关键要点
- 规格驱动开发（SDD） 是凭感觉编程的严谨替代方案，使用清晰的 Markdown 规格说明来指导编程智能体
- 三大核心优势：通过微小的规格修改控制大规模代码变更、消除上下文衰减、提高意图还原度
- 工作流：编写项目章程（任务愿景、技术栈、路线图），然后通过计划-实现-验证循环迭代开发功能
课程概览
欢迎来到基于编程智能体的规格驱动开发（Spec-Driven Development with Coding Agents）——一门由 JetBrains 与 DeepLearning.AI 联合推出的短课程。
讲师
Paul Everitt — JetBrains 开发者布道师。Paul 在开发者工具、Python Web 框架和 AI 辅助开发工作流方面拥有丰富的经验。
你将学到什么
本课程教你如何通过编写清晰的规格说明来指导编程智能体，从而将工程严谨性带入 AI 辅助编程中。你将学习：
- 如何编写项目章程（constitution），定义任务愿景、技术栈和路线图
- 如何编写功能规格说明（feature specs），隔离功能范围并实现计划-实现-验证循环
- 如何将 SDD 应用于全新项目（greenfield）
- 如何将 SDD 应用于传统项目（brownfield）（已有遗留代码库）
- 如何编写自己的**智能体技能（agent skills）**来自动化规格驱动工作流
为什么选择规格驱动开发？
规格驱动开发是目前在使用智能体编程助手构建严肃应用程序时最好的工作流。你不再需要手动编写代码，而是将注意力集中在写下智能体尚不具备的上下文信息上——给你的编程智能体提供一个 Markdown 文件或一段长提示词，准确说明要构建的内容。
课程大纲
- 简介 — SDD 优势与工作流概述
- 为什么选择规格驱动开发？ — 凭感觉编程 vs. SDD，用工程严谨性驾驭智能体
- 项目章程 — 为全新项目和传统项目编写项目级别的章程
- 功能规格说明 — 编写功能级别的规格说明，实现计划-实现-验证循环
- SDD 实践 — 在真实项目中的动手应用
- 智能体技能 — 用自定义智能体技能自动化你的 SDD 工作流