by zgsm-sangfor
Aggregates and indexes curated AI coding resources—MCP servers, skills, rules, and prompts—enabling one‑command search and installation across multiple AI coding platforms.
Coding Hub collects, cleans, evaluates, and presents over 3,900 carefully selected development resources. It unifies MCP servers, agent skills, coding rules, and prompts into a single searchable catalog, allowing AI agents and developers to find and install the right tool with a single command.
install.sh (macOS/Linux) or install.ps1 (Windows) script. /coding-hub:search <keyword> on Claude Code or /coding-hub-search <keyword> on Costrict CLI, to retrieve matching resources./coding-hub:install <name> or /coding-hub-install <name>) to add it to the current environment.Do I need to clone the repository? No. The installer fetches only the required skill and command files from the raw GitHub URLs.
Can I install resources on multiple platforms at once? Run the installer separately on each platform; the scripts detect the environment via injected variables.
What if a resource fails to install? The install command reports the error and provides the failing URL. Verify network access or try reinstalling.
How often is the catalog refreshed? Weekly via GitHub Actions; freshness scores are updated for each entry.
Is there a way to contribute new resources?
Yes – submit a pull request adding a properly formatted JSON entry under catalog/ and follow the schema in catalog/schema.json.
Where are the APIs hosted?
Static JSON endpoints are served from GitHub Pages, e.g., https://zgsm-sangfor.github.io/costrict-coding-hub/api/v1/search-index.json.
AI Coding Agent 越来越强,但找到合适的 MCP Server、Skill、Rule 仍然是碎片化的。
Coding Hub 从多类上游源自动聚合、清洗、评估开发资源,让你和你的 Agent 一条命令就能搜索和安装。
一条命令安装,指定你的平台:
macOS / Linux:
# Costrict CLI(在项目根目录执行)
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform costrict
# VSCode Costrict 插件
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform vscode-costrict
# Claude Code
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform claude-code
# Opencode(在项目根目录执行)
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform opencode
Windows (PowerShell):
# Costrict CLI
irm https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.ps1 | iex
# 指定平台(如自动检测失败)
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.ps1))) -Platform costrict
安装完成后试试:
# Claude Code
/coding-hub:search typescript
# Opencode / Costrict CLI / VSCode Costrict (Roo Code)
/coding-hub-search typescript
或者复制以下 prompt 丢给你的 AI Agent:
你是一个自动化安装助手。请访问以下 URL 并阅读其中的 "For Agents" 部分,
严格按照步骤完成 coding-hub 技能的安装。
不要 clone 整个仓库,只需要读取这个 raw URL 的内容:
https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/README.md
根据你所在的平台(Claude Code / Opencode / Costrict CLI / VSCode Costrict),
执行对应的安装命令。安装完成后告诉我结果。
从 3907+ 资源中按使用场景精选。安装后用
/coding-hub:search搜索完整索引,或/coding-hub:recommend获取项目级推荐。
Anthropic 官方社区精选Rules 2.1Rules 2.1wonderful-prompts社区精选antigravity-skillsRules 2.1CursorRulesprompts.chatAnthropic 官方社区精选精选Rules 2.1wonderful-promptsAnthropic 官方Anthropic 官方精选Rules 2.1prompts.chatAnthropic 官方Anthropic 官方Rules 2.1Rules 2.1精选davila7/claude-code-templatesdavila7/claude-code-templates精选Rules 2.1wonderful-promptsAnthropic 官方Anthropic 官方Rules 2.1Rules 2.1wonderful-prompts图例:🔌 MCP Server · 🎯 Skill · 📋 Rule · 💡 Prompt
| 类型 | 数量 | 说明 |
|---|---|---|
| MCP Server | 1629 | Model Context Protocol 服务器 |
| Prompt | 524 | 开发者专用 Prompt |
| Rule | 236 | 编码规范 / AI 辅助规则 |
| Skill | 1518 | Agent Skill 扩展 |
数据来源:由同步脚本从多类上游自动聚合;其中 Skills 的 Tier 2 还会通过 Registry 动态发现社区仓库。每周通过 GitHub Actions 同步,并发布到 GitHub Pages CDN。
| 上游 | 来源 |
|---|---|
| MCP | awesome-mcp-servers · Awesome-MCP-ZH · mcp.so |
| Skills | Tier 1: anthropics/skills · Ai-Agent-Skills · antigravity-awesome-skills · ai-agents-publicTier 2: awesome-repo-configs / skill_repos.json 动态发现社区仓库 · awesome-openclaw-skills · openclaw/skillsTier 3: catalog/skills/curated.json |
| Rules | awesome-cursorrules · rules-2.1-optimized |
| Prompts | prompts.chat · wonderful-prompts |
为了尽量保证结果可用,我们采用的是“分来源清洗 + 统一评估”的方式:不同类型会先按各自来源特点做基础筛选,再进入统一的评分、治理和持续维护流程。
第一层:来源侧清洗
mcp.so seed + awesome-mcp-servers + Awesome-MCP-ZH 聚合;对 awesome 列表里的 GitHub 仓库会补抓元数据,并保留基础活跃度筛选。skill_repos.json Registry 与 OpenClaw 发现候选,再过滤 spam、非 coding 分类和聚合仓库,并按确定性分数取 TOP 300。awesome-cursorrules 与 rules-2.1-optimized 的规则目录和 .mdc 文件,不额外套用统一的 star 门槛。prompts.chat 只保留面向开发者或命中 coding 关键词的条目;wonderful-prompts 只提取“编程”章节。第二层:分类来源与去重策略
| 类型 | 筛选策略 |
|---|---|
| MCP Servers | mcp.so seed > Awesome-MCP-ZH > awesome-mcp-servers 三源合并,按 GitHub URL(source_url)去重;必要时补抓 README 中的 mcpServers 配置 |
| Skills | Tier 1:官方 / 高质量来源经基础清洗后收录;Tier 2:Registry discovery + OpenClaw 候选,经确定性评分筛到 TOP 300;Tier 3:curated.json 作为最低优先级补充 |
| Rules | awesome-cursorrules + rules-2.1-optimized 双源聚合;在 merge 阶段按 id 去重 |
| Prompts | prompts.chat + wonderful-prompts 双源聚合;来源脚本先做 coding 相关过滤,merge 阶段按 id 去重 |
第三层:统一富化与评分治理
merge_index.py 会先加载各类型 index.json 与 curated.json,按照 Tier 1 > Tier 2 > Tier 3 的优先级去重。unified_enrichment.py)统一补齐 coding_relevance、content_quality、specificity、source_trust、confidence 等信号;有 LLM 或既有评估结果时优先复用,否则回落到启发式评分。scoring_governor.py)按类型权重计算 final_score,并写入 accept / review / reject 决策;health_scorer.py 再基于 popularity / freshness / quality / installability 生成健康度分数用于排序。持续维护
mcp / skill 条目会记录 added_at,表示首次进入 catalog 的日期catalog/index.json 保留顶层兼容字段,同时新增 evaluation 子对象承载统一评分与收录原因catalog/maintenance/incremental_recrawl_candidates.json 保存达到阈值的增量复抓候选,incremental_recrawl_state.json 保存去重/冷却状态catalog/mcp/crawl_state.json 与 catalog/skills/.repo_cache.json 继续负责各自来源的增量同步,不直接替代 catalog 生命周期治理支持四个 AI Coding 平台,命令格式略有差异:
| Claude Code | Costrict | Opencode | VSCode Costrict (Roo Code) | |
|---|---|---|---|---|
| 搜索 | /coding-hub:search <kw> [type:mcp] |
/coding-hub-search <kw> [type:mcp] |
/coding-hub-search <kw> [type:mcp] |
/coding-hub-search <kw> [type:mcp] |
| 浏览 | /coding-hub:browse [cat] |
/coding-hub-browse [cat] |
/coding-hub-browse [cat] |
/coding-hub-browse [cat] |
| 推荐 | /coding-hub:recommend |
/coding-hub-recommend |
/coding-hub-recommend |
/coding-hub-recommend |
| 安装 | /coding-hub:install <name> |
/coding-hub-install <name> |
/coding-hub-install <name> |
/coding-hub-install <name> |
| 卸载 | /coding-hub:uninstall <name> |
/coding-hub-uninstall <name> |
/coding-hub-uninstall <name> |
/coding-hub-uninstall <name> |
| 更新 | /coding-hub:update |
/coding-hub-update |
/coding-hub-update |
/coding-hub-update |
| Claude Code | Costrict | VSCode Costrict | Opencode | |
|---|---|---|---|---|
| Skill 路径(全局) | ~/.claude/skills/coding-hub/ |
~/.costrict/skills/coding-hub/ |
~/.costrict/skills/coding-hub/ |
~/.opencode/skills/coding-hub/ |
| Commands 路径 | 同上(全局) | .costrict/coding-hub/commands/(项目级) |
~/.roo/commands/(全局) |
.opencode/command/(项目级) |
| 命令分隔符 | : |
- |
- |
- |
install.sh 即可~/.roo/commands/,通过 /coding-hub-update 自动下载install.sh 即可costrict-coding-hub/
├── install.sh # 一键安装脚本(macOS/Linux,curl | bash)
├── install.ps1 # 一键安装脚本(Windows,irm | iex)
├── catalog/ # 资源索引(数据层)
│ ├── index.json # 合并后的完整索引(3900+ 条)
│ ├── search-index.json # 轻量搜索索引(仅搜索字段,~2MB)
│ ├── schema.json # 条目 schema 定义
│ ├── mcp/ # MCP Server 源数据(含 added_at 生命周期字段)
│ ├── skills/ # Skill 源数据(含 added_at 生命周期字段)
│ ├── rules/ # Rule 源数据
│ ├── prompts/ # Prompt 源数据
│ └── maintenance/ # 增量复抓候选与状态
│
├── docs/api/ # GitHub Pages 静态 API(CI 生成,不提交)
│ └── v1/ # API v1
│ ├── search-index.json # 搜索索引副本
│ ├── {type}/index.json # 各类型轻量索引
│ └── {type}/{id}.json # 单条完整数据(~1-2KB)
│
├── platforms/ # 各平台 Skill + 子命令
│ ├── claude-code/ # Claude Code 格式(命令分隔符 `:`)
│ ├── opencode/ # Opencode 格式
│ ├── costrict/ # Costrict 格式
│ └── vscode-costrict/ # VSCode Costrict 插件格式
│
├── scripts/ # 数据同步与生成脚本
│ ├── sync_mcp.py # 从上游同步 MCP
│ ├── sync_skills.py # 从上游同步 Skills
│ ├── sync_rules.py # 从上游同步 Rules
│ ├── sync_prompts.py # 从上游同步 Prompts
│ ├── crawl_mcp_so.py # 增量抓取 mcp.so
│ ├── merge_index.py # 合并生成 index.json(去重→富化→评分→生命周期)
│ ├── generate_pages.py # 生成 GitHub Pages 静态 API(按条目拆分)
│ ├── update_readme.py # 自动更新 README 统计数字
│ ├── enrichment_orchestrator.py # 富化调度
│ ├── unified_enrichment.py # Layer 2: source_trust, confidence
│ ├── llm_evaluator.py # LLM 质量评估
│ ├── scoring_governor.py # 评分治理: final_score 加权
│ ├── health_scorer.py # 健康评分四维信号
│ ├── catalog_lifecycle.py # 生命周期字段与增量复抓
│ └── utils.py # 公共工具函数
│
├── .github/workflows/ # CI/CD
│ ├── sync.yml # 每周自动同步上游资源
│ ├── deploy-pages.yml # 同步后自动部署 GitHub Pages 静态 API
│ ├── test.yml # PR 测试
│ └── validate-pr.yml # PR 校验
│
└── openspec/ # 变更管理(OpenSpec)
AI Agent 请阅读本节完成自动安装。 请严格按照以下步骤执行,不要跳过任何一步。
安装脚本支持自动检测平台,直接执行即可:
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash
Windows (PowerShell):
irm https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.ps1 | iex
脚本通过以下进程级环境变量自动识别平台(这些变量由各平台进程启动时注入,不会互相污染):
| 环境变量 | 平台 |
|---|---|
COSTRICT_CALLER=vscode |
VSCode Costrict 插件 |
COSTRICT_RUNNING=1 |
Costrict CLI |
CLAUDECODE=1 |
Claude Code |
OPENCODE=1 |
Opencode |
如果自动检测失败,可以手动指定:
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform <platform>
Windows (PowerShell):
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.ps1))) -Platform <platform>
示例:
# Claude Code (macOS/Linux)
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform claude-code
# Opencode(在项目根目录执行)
curl -fsSL https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.sh | bash -s -- --platform opencode
# Claude Code (Windows)
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/install.ps1))) -Platform claude-code
根据你的平台,手动执行对应命令: Costrict CLI:(Skill 全局 + Commands 项目级)
mkdir -p ~/.costrict/skills/coding-hub
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/costrict/skills/coding-hub/SKILL.md" -o ~/.costrict/skills/coding-hub/SKILL.md
mkdir -p .costrict/coding-hub/commands
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/costrict/commands/coding-hub/coding-hub-${cmd}.md" -o .costrict/coding-hub/commands/coding-hub-${cmd}.md
done
VSCode Costrict 插件:(Skill 全局 + Commands 全局)
mkdir -p ~/.costrict/skills/coding-hub
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/vscode-costrict/skills/coding-hub/SKILL.md" -o ~/.costrict/skills/coding-hub/SKILL.md
mkdir -p ~/.roo/commands
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/vscode-costrict/commands/coding-hub/coding-hub-${cmd}.md" -o ~/.roo/commands/coding-hub-${cmd}.md
done
Roo Code 支持原生 slash commands,命令文件安装到全局
~/.roo/commands/。
Claude Code:(Skill + Commands 全部安装到全局目录)
mkdir -p ~/.claude/skills/coding-hub
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/claude-code/skills/coding-hub/SKILL.md" -o ~/.claude/skills/coding-hub/SKILL.md
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/claude-code/commands/coding-hub/${cmd}.md" -o ~/.claude/skills/coding-hub/${cmd}.md
done
Opencode:(Skill 全局 + Commands 项目级)
mkdir -p ~/.opencode/skills/coding-hub
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/opencode/skills/coding-hub/SKILL.md" -o ~/.opencode/skills/coding-hub/SKILL.md
mkdir -p .opencode/command
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/platforms/opencode/command/coding-hub-${cmd}.md" -o .opencode/command/coding-hub-${cmd}.md
done
执行搜索命令验证安装是否成功:
# Costrict CLI / Opencode
/coding-hub-search typescript
# VSCode Costrict 插件
在插件聊天中发送: "用 coding-hub 搜索 typescript"
# Claude Code
/coding-hub:search typescript
预期结果: 返回一个包含搜索结果的 Markdown 表格。如果返回了表格,安装成功。如果提示命令不存在,请检查第 1 步的安装是否正确执行。
安装成功后,以下命令可用(命令分隔符根据平台不同为 : 或 -):
| 命令 | 说明 |
|---|---|
search <关键词> [type:mcp|skill|rule|prompt] |
搜索资源,可按类型筛选 |
browse [分类] [type:...] |
浏览分类 |
recommend [type:...] |
基于当前项目技术栈推荐 |
install <名称> |
安装资源 |
uninstall <名称> |
卸载资源 |
update |
更新到最新版本 |
搜索/浏览/推荐使用轻量搜索索引(~2MB),安装使用单条 API(~1KB),均通过 GitHub Pages CDN 分发:
| 用途 | URL |
|---|---|
| 搜索索引 | https://zgsm-sangfor.github.io/costrict-coding-hub/api/v1/search-index.json |
| 单条 API | https://zgsm-sangfor.github.io/costrict-coding-hub/api/v1/{type}/{id}.json |
| 类型索引 | https://zgsm-sangfor.github.io/costrict-coding-hub/api/v1/{type}/index.json |
| 全量索引(fallback) | https://raw.githubusercontent.com/zgsm-sangfor/costrict-coding-hub/main/catalog/index.json |
索引是 JSON 数组,每个条目包含 id, name, type(mcp/skill/rule/prompt), description, source_url, stars, category, tags, tech_stack, install。
如果你喜欢 Coding Hub 的理念,不妨试试我们的旗舰产品 Costrict —— 一个更强大的 AI Coding Agent 平台:
欢迎通过 PR 向 catalog/ 下对应类型目录添加精选资源。提交前请确保:
source_url、description、tagscatalog/schema.json 定义的数据格式Coding Hub 是一个资源索引聚合项目,所有收录的 MCP Server、Skill、Rule、Prompt 均来自第三方开源仓库,版权归各自作者所有。本项目仅提供索引和安装便利,不对第三方资源的安全性、可用性、准确性或合规性做任何保证。
使用本项目收录的任何资源所产生的风险由用户自行承担。建议在使用前审查资源的源代码和许可证。如发现安全问题或侵权内容,请通过 Issue 反馈,我们会及时处理。
本项目以 MIT License 发布。
Please log in to share your review and rating for this MCP.
Explore related MCPs that share similar capabilities and solve comparable challenges
by modelcontextprotocol
A Model Context Protocol server for Git repository interaction and automation.
by zed-industries
A high‑performance, multiplayer code editor designed for speed and collaboration.
by modelcontextprotocol
Model Context Protocol Servers
by modelcontextprotocol
A Model Context Protocol server that provides time and timezone conversion capabilities.
by cline
An autonomous coding assistant that can create and edit files, execute terminal commands, and interact with a browser directly from your IDE, operating step‑by‑step with explicit user permission.
by upstash
Provides up-to-date, version‑specific library documentation and code examples directly inside LLM prompts, eliminating outdated information and hallucinated APIs.
by daytonaio
Provides a secure, elastic infrastructure that creates isolated sandboxes for running AI‑generated code with sub‑90 ms startup, unlimited persistence, and OCI/Docker compatibility.
by continuedev
Enables faster shipping of code by integrating continuous AI agents across IDEs, terminals, and CI pipelines, offering chat, edit, autocomplete, and customizable agent workflows.
by github
Connects AI tools directly to GitHub, enabling natural‑language interactions for repository browsing, issue and pull‑request management, CI/CD monitoring, code‑security analysis, and team collaboration.