经 AI Skill Hub 精选评估,SpecFlow 获评「强烈推荐」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 8.5 分,适合有一定技术背景的用户使用。
SpecFlow 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
SpecFlow 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:go install(推荐) go install github.com/Bingordinary/SpecFlow@latest # 方式二:从源码编译 git clone https://github.com/Bingordinary/SpecFlow cd SpecFlow go build -o specflow . # 方式三:下载预编译二进制 # 访问 Releases 页面下载对应平台二进制文件 # https://github.com/Bingordinary/SpecFlow/releases
# 查看帮助 specflow --help # 基本运行 specflow [options] <input> # 详细使用说明请查阅文档 # https://github.com/Bingordinary/SpecFlow
# specflow 配置说明 # 查看配置选项 specflow --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export SPECFLOW_CONFIG="/path/to/config.yml"
<p> <img alt="spec-driven" src="https://img.shields.io/badge/spec-driven-111111?style=for-the-badge&labelColor=111111&color=2F855A"> <img alt="unit-governed" src="https://img.shields.io/badge/unit-governed-111111?style=for-the-badge&labelColor=111111&color=1F6FEB"> <img alt="agent-runtime-ready" src="https://img.shields.io/badge/agent-runtime%20ready-111111?style=for-the-badge&labelColor=111111&color=C2410C"> <img alt="human-and-ai" src="https://img.shields.io/badge/human%20%2B%20AI-collaboration-111111?style=for-the-badge&labelColor=111111&color=7C3AED"> </p>
English · 简体中文
Add To Your Repository · Quick Start · Adoption Modes · Core Concepts · Standard Commands · Development Workflow · Reader · Advanced
---
specFlow makes AI-assisted development feel like engineering again: instead of letting requirements dissolve into chat logs, code diffs, and personal memory, it gives every governed unit a current truth, a next truth, and a clear path from idea to verified change. Humans and agents can move fast together while the repository still knows what is true, what is changing, and what is ready to ship.
It is not a fixed business template, and it does not force every team to write the same documents. It is an engineering collaboration skeleton: requirements enter repository truth first, then planning, implementation, verification, and promotion follow that truth.
If you used the installer, init has already run. After manual setup, once the binaries are in place and specflow/ is in your repository, run from the project root:
<specflow-binary> init
<specflow-binary> means the platform-matching specflowctl executable under specflow/tooling/bin/. See tooling/README.md for exact filenames.
init installs the basic structure:
AGENTS.md, GEMINI.md, and CLAUDE.mddocs/specs/After this step, choose an Adoption Mode. init prepares the shared skeleton; it does not require you to run the whole lifecycle immediately.
For daily work, use standard commands with the module name you want to work on:
unit_new:{module_name}
unit_check:{module_name}
unit_fork:{module_name}
unit_verify:{module_name}
When unsure, fall back to natural language:
I want to add rate limiting to auth, but I am not sure what should move first. Read current project truth and tell me the next step.
The agent reads the installed entry files and current repository truth, then decides which command to enter, whether to write Spec truth, check a boundary, or ask a required clarification.
You can start small. Installing specFlow does not commit a project to promotion, stable verification, governance review, or full lifecycle use.
| Mode | Use When | What It Allows |
|---|---|---|
reader-only | You want visibility before changing process | Start specflow-reader, inspect state and truth, make no lifecycle writes |
implementation-only | The request fits already-written formal truth | Use natural language for a code or test change; stop if truth, boundary, acceptance, rule, or ownership must change |
single-unit-trial | You want to try specFlow on one unit | Govern one named unit through the needed steps while leaving the rest of the repo outside specFlow |
unit-check-only | You only want to test whether a Spec is a good requirement | Run unit_check:{unit} and stop after pass, blocked, or fix-required evidence |
The formal contract is specflow/framework/core/adoption_modes.md. These modes are entry choices, not new lifecycle states, process schema, or CLI mode switches. Promotion, stable verification, and governance review remain explicit later choices, not default requirements for these starts.
Example reader-only start:
<specflow-reader-binary> --repo-root . --addr 127.0.0.1:17863
Example implementation-only request:
Make the existing retry test less flaky without changing the documented behavior. If this needs truth changes, stop and tell me the smallest specFlow step.
Example single-unit trial:
Use specFlow only for the payment_retry unit for now. Do not promote or enter governance review unless I ask.
Example unit-check-only request:
Run unit_check:payment_retry and stop after the check result. Do not plan or implement yet.
高质量的开源AI工作流项目
该工具未明确声明开源协议,商业使用前请联系原作者确认授权范围,避免侵权风险。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
AI Skill Hub 点评:SpecFlow 的核心功能完整,质量优秀。对于自动化工程师和运维人员来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。
| 原始名称 | SpecFlow |
| 原始描述 | 开源AI工作流:`specFlow` is a module-oriented, spec-driven development paradigm for teams that。⭐8 · Go |
| Topics | AI工作流Go |
| GitHub | https://github.com/Bingordinary/SpecFlow |
| 语言 | Go |
收录时间:2026-05-31 · 更新时间:2026-05-31 · License:未公布 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端