feat: package Arthas skills for agent installers

.agents/plugins/marketplace.json

@@ -0,0 +1,20 @@
+{
+  "name": "arthas-plugins",
+  "interface": {
+    "displayName": "Arthas Plugins"
+  },
+  "plugins": [
+    {
+      "name": "arthas",
+      "source": {
+        "source": "local",
+        "path": "./plugins/arthas"
+      },
+      "policy": {
+        "installation": "AVAILABLE",
+        "authentication": "ON_INSTALL"
+      },
+      "category": "Developer Tools"
+    }
+  ]
+}

README.md

@@ -87,6 +87,7 @@ You can enter its interactive interface by executing `as.sh`, or execute `as.sh
 * [Quick start](https://arthas.aliyun.com/doc/en/quick-start.html)
 * [Advanced usage](https://arthas.aliyun.com/doc/en/advanced-use.html)
 * [Commands](https://arthas.aliyun.com/doc/en/commands.html)
+* [AI agent skills installation](plugins/arthas/README.md)
 * [WebConsole](https://arthas.aliyun.com/doc/en/web-console.html)
 * [Docker](https://arthas.aliyun.com/doc/en/docker.html)
 * [Arthas Spring Boot Starter](https://arthas.aliyun.com/doc/en/spring-boot-starter.html)

README_CN.md

@@ -76,6 +76,7 @@ curl -L https://arthas.aliyun.com/install.sh | sh
 * [快速入门](https://arthas.aliyun.com/doc/quick-start.html)
 * [进阶使用](https://arthas.aliyun.com/doc/advanced-use.html)
 * [命令列表](https://arthas.aliyun.com/doc/commands.html)
+* [AI Agent Skills 安装](plugins/arthas/README_CN.md)
 * [WebConsole](https://arthas.aliyun.com/doc/web-console.html)
 * [Docker](https://arthas.aliyun.com/doc/docker.html)
 * [Arthas Spring Boot Starter](https://arthas.aliyun.com/doc/spring-boot-starter.html)

plugins/arthas/.codex-plugin/plugin.json

@@ -0,0 +1,28 @@
+{
+  "name": "arthas",
+  "version": "0.1.0",
+  "description": "Guided Arthas workflows for diagnosing Java and JVM issues.",
+  "author": {
+    "name": "Alibaba Arthas",
+    "url": "https://github.com/alibaba/arthas"
+  },
+  "homepage": "https://arthas.aliyun.com/",
+  "repository": "https://github.com/alibaba/arthas",
+  "license": "Apache-2.0",
+  "keywords": [
+    "arthas",
+    "java",
+    "jvm",
+    "diagnostics"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "Arthas Diagnostics",
+    "shortDescription": "Diagnose Java and JVM issues with Arthas.",
+    "longDescription": "Use guided Arthas workflows to investigate CPU spikes, EagleEye trace IDs, and Spring context or bean issues.",
+    "developerName": "Alibaba Arthas",
+    "category": "Developer Tools",
+    "capabilities": [],
+    "defaultPrompt": "Help me diagnose a Java application with Arthas."
+  }
+}

plugins/arthas/README.md

@@ -0,0 +1,89 @@
+# Arthas AI Agent Skills
+
+Arthas provides reusable skills that guide AI coding agents through common Java and JVM diagnostic workflows.
+
+[中文说明](README_CN.md)
+
+## Available skills
+
+| Skill | Purpose |
+| --- | --- |
+| `arthas` | Routes a diagnostic request to the most relevant Arthas workflow. |
+| `arthas-cpu-high` | Investigates high CPU usage from JVM state to hot threads and methods. |
+| `arthas-eagleeye-traceid` | Captures an EagleEye trace ID with bounded `watch` or `trace` commands. |
+| `arthas-springcontext-issues-resolve` | Investigates Spring contexts, beans, and configuration values. |
+
+The repository uses one canonical skill tree at `plugins/arthas/skills/`. The same files can be installed as a Codex plugin, with the cross-agent `skills` CLI, or with the GitHub `gh-skill` extension.
+
+## Option 1: Codex plugin
+
+This option installs all four skills together and makes the plugin available in Codex's plugin directory.
+
+Add the repository marketplace with a sparse checkout, then install the plugin:
+
+```bash
+codex plugin marketplace add alibaba/arthas \
+  --sparse .agents/plugins \
+  --sparse plugins/arthas
+codex plugin add arthas@arthas-plugins
+```
+
+Verify the installation:
+
+```bash
+codex plugin list --json
+```
+
+You can also open the Plugins Directory in the ChatGPT desktop app, choose **Arthas Plugins**, and install **Arthas Diagnostics**.
+
+See the official [plugin packaging guide](https://developers.openai.com/plugins/build/plugins) and [Codex plugin commands](https://learn.chatgpt.com/docs/developer-commands#codex-plugin) for more details.
+
+## Option 2: `skills` CLI
+
+Use this option when you want the same skills across Codex, Claude Code, Cursor, or another supported agent.
+
+List the skills without installing them:
+
+```bash
+npx skills add alibaba/arthas --list
+```
+
+Install every Arthas skill for Codex in the current project:
+
+```bash
+npx skills add alibaba/arthas --agent codex --skill '*' --yes
+```
+
+Add `--global` to make them available to Codex in every project. To install only one workflow, replace `'*'` with its skill name, for example `arthas-cpu-high`.
+
+Update project-scoped installations with:
+
+```bash
+npx skills update --project --yes
+```
+
+## Option 3: GitHub `gh-skill`
+
+Use this option when you prefer GitHub-native discovery, preview, version pinning, and updates.
+
+Browse the available skills interactively:
+
+```bash
+gh skill install alibaba/arthas
+```
+
+Preview and install one workflow directly:
+
+```bash
+gh skill preview alibaba/arthas arthas-cpu-high
+gh skill install alibaba/arthas arthas-cpu-high \
+  --agent codex \
+  --scope user
+```
+
+Check for updates without changing local files, then apply all available updates:
+
+```bash
+gh skill update --dry-run
+gh skill update --all
+```

plugins/arthas/README_CN.md

@@ -0,0 +1,89 @@
+# Arthas AI Agent Skills
+
+Arthas 提供了一组可复用的 Skill，帮助 AI 编码 Agent 按规范流程诊断常见的 Java 与 JVM 问题。
+
+[English](README.md)
+
+## 可用 Skill
+
+| Skill | 用途 |
+| --- | --- |
+| `arthas` | 根据问题描述选择最合适的 Arthas 诊断流程。 |
+| `arthas-cpu-high` | 从 JVM 概况、热点线程到热点方法，排查 CPU 飙高。 |
+| `arthas-eagleeye-traceid` | 使用有限次数的 `watch` 或 `trace` 获取 EagleEye traceId。 |
+| `arthas-springcontext-issues-resolve` | 排查 Spring Context、Bean 和配置项问题。 |
+
+仓库以 `plugins/arthas/skills/` 作为唯一 Skill 源目录。同一份文件既可以作为 Codex Plugin 安装，也兼容跨 Agent 的 `skills` CLI 和 GitHub `gh-skill` 扩展。
+
+## 方式一：Codex Plugin
+
+这种方式会一次安装全部四个 Skill，并在 Codex 的插件目录中展示该插件。
+
+先用稀疏检出添加仓库 Marketplace，再安装插件：
+
+```bash
+codex plugin marketplace add alibaba/arthas \
+  --sparse .agents/plugins \
+  --sparse plugins/arthas
+codex plugin add arthas@arthas-plugins
+```
+
+检查安装结果：
+
+```bash
+codex plugin list --json
+```
+
+也可以在 ChatGPT 桌面端打开插件目录，选择 **Arthas Plugins**，再安装 **Arthas Diagnostics**。
+
+更多信息请参考官方的 [Plugin 打包指南](https://developers.openai.com/plugins/build/plugins)和 [Codex Plugin 命令](https://learn.chatgpt.com/docs/developer-commands#codex-plugin)。
+
+## 方式二：`skills` CLI
+
+需要在 Codex、Claude Code、Cursor 或其他兼容 Agent 之间复用 Skill 时，可以选择这种方式。
+
+只查看仓库中的 Skill，不执行安装：
+
+```bash
+npx skills add alibaba/arthas --list
+```
+
+为当前项目的 Codex 安装全部 Arthas Skill：
+
+```bash
+npx skills add alibaba/arthas --agent codex --skill '*' --yes
+```
+
+追加 `--global` 可以让它们在 Codex 的所有项目中生效。只安装一个诊断流程时，把 `'*'` 替换成对应 Skill 名称，例如 `arthas-cpu-high`。
+
+更新项目级安装：
+
+```bash
+npx skills update --project --yes
+```
+
+## 方式三：GitHub `gh-skill`
+
+希望使用 GitHub 原生的发现、预览、版本锁定和更新能力时，可以选择这种方式。
+
+交互式浏览可用 Skill：
+
+```bash
+gh skill install alibaba/arthas
+```
+
+直接预览并安装一个诊断流程：
+
+```bash
+gh skill preview alibaba/arthas arthas-cpu-high
+gh skill install alibaba/arthas arthas-cpu-high \
+  --agent codex \
+  --scope user
+```
+
+先只读检查更新，再应用全部可用更新：
+
+```bash
+gh skill update --dry-run
+gh skill update --all
+```

skills/cpu-high/SKILL.md → plugins/arthas/skills/arthas-cpu-high/SKILL.md

@@ -1,6 +1,7 @@
 ---
 name: arthas-cpu-high
 description: 排查 JVM / 应用 CPU 飙高（线程定位 + 代码路径分析）
+license: Apache-2.0
 ---
 
 # JVM CPU 飙高排查指南（Arthas）
@@ -36,4 +37,3 @@ description: 排查 JVM / 应用 CPU 飙高（线程定位 + 代码路径分析
 - 现象与证据：dashboard 摘要 + topN 线程堆栈关键片段
 - 初步结论：CPU 主要消耗在什么类型的逻辑（计算/锁/GC/日志等）
 - 下一步：建议进一步 trace/watch 的目标方法（给出类名+方法名的精确范围），或建议用户提供主包名/关键接口信息以继续收敛
-

skills/eagleeye-traceid/SKILL.md → plugins/arthas/skills/arthas-eagleeye-traceid/SKILL.md

@@ -1,6 +1,7 @@
 ---
 name: arthas-eagleeye-traceid
 description: 使用 Arthas 的 watch/trace 获取 EagleEye traceId / 获取请求的 traceId
+license: Apache-2.0
 ---
 
 # 获取 EagleEye traceId（Arthas）
@@ -77,4 +78,3 @@ trace <类全名> <方法名> -n 5
 ## 扩展
 
 - 其他分布式追踪系统的 traceId、或 ThreadLocal 里的值，也可以用类似方式在 `watch` 的 OGNL 表达式中读取。
-

skills/spring-context/SKILL.md → plugins/arthas/skills/arthas-springcontext-issues-resolve/SKILL.md

@@ -1,6 +1,7 @@
 ---
 name: arthas-springcontext-issues-resolve
 description: 排查 Spring ApplicationContext / Bean / 配置注入等问题
+license: Apache-2.0
 ---
 
 # Spring Context / Bean 排查指南
@@ -97,4 +98,4 @@ vmtool --action getInstances --className org.springframework.context.support.Abs
 
 ```bash
 vmtool --action getInstances --className org.springframework.context.support.AbstractApplicationContext -l 1 --express '#ctx=instances[0], #bf=#ctx.getBeanFactory(), #bd=#bf.getBeanDefinition("fooService")'
-```
+```

skills/SKILL.md → plugins/arthas/skills/arthas/SKILL.md

@@ -1,6 +1,7 @@
 ---
 name: arthas
 description: arthas 诊断 java应用，jvm问题 skill
+license: Apache-2.0
 ---
 
 # Arthas 诊断 Skill
@@ -20,7 +21,7 @@ Arthas 是 Java 应用在线诊断工具，本 Skill 包含多个子场景的诊
 
 ### 1. CPU 飙高排查
 
-**文件：** `cpu-high/SKILL.md`
+**文件：** `../arthas-cpu-high/SKILL.md`
 
 适用场景：机器 CPU 飙高、应用响应变慢、负载异常升高。
 
@@ -35,7 +36,7 @@ Arthas 是 Java 应用在线诊断工具，本 Skill 包含多个子场景的诊
 
 ### 2. 获取 EagleEye traceId
 
-**文件：** `eagleeye-traceid/SKILL.md`
+**文件：** `../arthas-eagleeye-traceid/SKILL.md`
 
 适用场景：需要在不改代码的情况下，获取线上请求的 EagleEye traceId，用于关联日志 / 链路分析。
 
@@ -50,7 +51,7 @@ Arthas 是 Java 应用在线诊断工具，本 Skill 包含多个子场景的诊
 
 ### 3. Spring Context / Bean 排查
 
-**文件：** `spring-context/SKILL.md`
+**文件：** `../arthas-springcontext-issues-resolve/SKILL.md`
 
 适用场景：排查 Spring ApplicationContext / Bean / 配置注入等问题。