# cc-sdd — Spec-Driven Implementation Skills for Agents

> One-command installer that adds spec-driven SDLC skills (Kiro-style) and long-running implementation flows across multiple coding agents.

## Install

Save the content below to `.claude/skills/` or append to your `CLAUDE.md`:

## Quick Use

1. Install into a project:
   ```bash
   cd your-project
   npx cc-sdd@latest
   ```
2. Pick a target agent/language (examples):
   ```bash
   npx cc-sdd@latest --codex-skills --lang ja
   npx cc-sdd@latest --cursor-skills --lang zh-TW
   ```
3. Start the workflow inside your agent:
   - Run `/kiro-discovery <idea>`

## Intro

One-command installer that adds spec-driven SDLC skills (Kiro-style) and long-running implementation flows across multiple coding agents.

- **Best for:** teams who want long-running, reviewable, spec-driven delivery with repeatable phases and fewer context resets
- **Works with:** Node.js + npx, multiple coding agents; writes skills/workflow assets into your repo
- **Setup time:** 20–45 minutes

## Practical Notes

- Quant: the README calls out 17 skills across 8 agents and 13 languages—start with 1 agent + 1 language in production, then expand.
- Quant: measure cycle time per phase (discovery→requirements→design→tasks→impl) on 3 similar features before and after adoption.

---

## Make specs a contract, not a monolith

Spec-driven delivery works when you treat outputs as contracts:

- Requirements define acceptance criteria.
- Design defines boundaries.
- Tasks define interfaces and dependencies.

## A rollout recipe

1. Pick one feature that is “medium sized”.
2. Run discovery, then stop and review `brief.md` / `roadmap.md`.
3. Only then run the implementation flow.

## Guardrails that matter

- Keep feature flags for autonomous changes.
- Require an independent review pass.
- Use a failure log: every agent failure must become a checklist item or test.

### FAQ

**Q: Will this slow down small changes?**
A: Use it selectively: spec-driven shines on medium/large work; tiny fixes can skip full phases.

**Q: What is the biggest win?**
A: Less context thrash: phases are explicit and artifacts are resumable.

**Q: How do I keep quality high?**
A: Gate each phase: humans approve the contract before the next phase runs.

## Source & Thanks

> Source: https://github.com/gotalab/cc-sdd
> License: MIT
> GitHub stars: 3,324 · forks: 245

---

<!-- ZH -->

## 快速使用

1. 在项目里安装：
   ```bash
   cd your-project
   npx cc-sdd@latest
   ```
2. 选择目标 Agent/语言（示例）：
   ```bash
   npx cc-sdd@latest --codex-skills --lang ja
   npx cc-sdd@latest --cursor-skills --lang zh-TW
   ```
3. 在 agent 中启动流程：
   - 执行 `/kiro-discovery <idea>`

## 简介

一键安装规格驱动的 SDLC 技能包（Kiro 风格）：把 discovery→requirements→design→tasks→implementation 变成可长期运行的工作流，并适配多种编码 Agent。

- **适合谁：** 想把交付做成可长期运行、可复盘的规格驱动流程，并减少频繁“重讲上下文”的团队
- **可搭配：** Node.js + npx、多种编码 Agent；会把 skills/工作流产物写入你的仓库
- **准备时间：** 20–45 分钟

## 实战建议

- 量化建议：README 提到 8 个 agents、13 种语言、17 个 skills；生产环境先只选 1 个 agent + 1 种语言跑稳再扩展。
- 量化建议：在 3 个相似需求上对比 adoption 前后各阶段耗时（discovery→requirements→design→tasks→impl）。

## 把规格当作“契约”，而不是“巨型命令”

规格驱动的关键是把输出当作契约：

- requirements 负责验收标准。
- design 负责边界与约束。
- tasks 负责接口与依赖顺序。

## 一套好用的推广步骤

1. 选一个“中等规模”的需求试点。
2. 先跑 discovery，然后停下来 review `brief.md`/`roadmap.md`。
3. 通过后再启动实现流程。

## 真正重要的护栏

- 自主改动尽量放在 feature flag 后。
- 强制独立 review 通过。
- 建立失败日志：每次失败都要沉淀为清单项或测试用例。

### FAQ

**会不会拖慢小改动？**
答：选择性使用：中大型需求收益最大；小修小补不必强行走全流程。

**最大的收益是什么？**
答：减少上下文反复解释：阶段明确、产物可续跑。

**怎么保证质量？**
答：阶段闸门：下一阶段启动前由人类审核通过上一阶段契约。

## 来源与感谢

> Source: https://github.com/gotalab/cc-sdd
> License: MIT
> GitHub stars: 3,324 · forks: 245


---
Source: https://tokrepo.com/en/workflows/cc-sdd-spec-driven-implementation-skills-for-agents
Author: Skill Factory