inbox-setup

作者: microsoft

使用推荐的 VS Code 设置和 gh CLI 配置来设置 Inbox 代理

npx skills add https://github.com/microsoft/vscode-team-kit --skill inbox-setup

Inbox Agent Setup

Before running any setup steps, use #askQuestions to ask: "Some required settings are missing for the Inbox agent. Would you like me to configure them automatically?" with choices "Yes, set up now" and "Skip for now".

If the user skips, stop and let them proceed without setup.

Step 0: Detect existing memory

Before touching VS Code settings, check whether the user already has inbox memory from a previous setup (e.g. Copilot CLI):

  1. Run ls -la ~/.copilot/github-inbox-*.md 2>/dev/null to detect existing CLI memory/rules files.
  2. Try #memory view /memories/github-inbox-memory.md and /memories/github-inbox-rules.md to detect existing VS Code memory.
  3. Report what you found in plain language, e.g.:
    • "Found existing memory in ~/.copilot/: github-inbox-memory.md (84 lines), github-inbox-rules.md (16 rules)."
    • "No existing VS Code #memory files."
  4. If existing memory was found in EITHER location, use #askQuestions to ask which the inbox agent should use:
    • "Use existing ~/.copilot/ files (recommended — keep your rules and history)"
    • "Use VS Code #memory (start fresh)"
    • "Use VS Code #memory and import from ~/.copilot/ (copy once, then use #memory)"
  5. Record the choice in memory under a ## Backend section so future sessions don't re-prompt.

If no existing memory is found anywhere, default to #memory silently and continue.

Step 1: Check gh CLI

gh --version

If not installed, see the inbox-install-gh-cli skill.

Step 2: Configure VS Code settings

  1. Ask the user for their VS Code settings.json path if needed, or check common locations (~/Library/Application Support/Code/User/settings.json on macOS, ~/.config/Code/User/settings.json on Linux, %APPDATA%\Code\User\settings.json on Windows)
  2. Read settings using the read tool
  3. Check which of the required settings below are missing
  4. If ALL required settings are already present with correct values, skip this step
  5. If any are missing, use VS Code's editFile tool to insert ONLY the missing keys into the user's settings.json. This is safer than terminal writes for JSON files.

CRITICAL: settings.json contains the user's entire VS Code configuration. You MUST NOT remove any existing content. Only ADD the specific missing keys. If chat.tools.terminal.autoApprove already exists with some rules, merge the new rules into the existing object — don't replace it.

Required settings to add if missing:

{
  "chat.customAgentInSubagent.enabled": true,
  "github.copilot.chat.githubMcpServer.enabled": true,
  "chat.tools.terminal.enableAutoApprove": true,
  "chat.tools.terminal.autoApprove": {
    "/^gh api (?!.*--method (PATCH|PUT|POST|DELETE))/": true,
    "/^gh search /": true,
    "gh issue view": true,
    "gh pr view": true,
    "gh --version": true,
    "gh config set": true
  }
}

Tell the user what was added and why:

  • chat.customAgentInSubagent.enabled — Required. Enables sub-agent delegation (reviewer, triager, memory, investigator).
  • github.copilot.chat.githubMcpServer.enabled — Enables richer features: PR reviews, comments, labels, assignees.
  • chat.tools.terminal.enableAutoApprove + chat.tools.terminal.autoApprove — Auto-approves read-only gh commands.

Step 3: Enable auto-approve

After updating settings, tell the user:

"✅ Settings updated! One last step — I need to enable auto-approve so read-only commands don't prompt you every time.

👉 When you see the 'Run command?' prompt below, click the Allow ▾ dropdown and select 'Enable Auto Approve'.

This is a one-time opt-in. After this, read-only gh commands will run automatically."

Then run:

gh api /notifications --jq 'length'

This command matches the auto-approve rules. Wait for the user to click "Enable Auto Approve" before proceeding.

Step 4: Verify GitHub MCP server

Check if the GitHub MCP server is running. If github.copilot.chat.githubMcpServer.enabled was just added, tell the user:

"Make sure the GitHub MCP server is started and running. This gives you access to richer features like PR reviews, comments, and labels."

Step 5: Save setup status

After setup is complete, save to memory that setup has been completed:

Use the inbox-memory skill to save setup status:

Write to memory: "# Inbox Memory\n\n## Setup State\n- Setup completed on <date>\n- All required VS Code settings are present and verified\n"

If the memory file already exists, use delete first, then create with the merged content.

来自 microsoft 的更多技能

oss-growth
microsoft
OSS增长黑客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。在创建使用AzureAIAgentsProvider的持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应时使用。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上设置AI Runway——从裸集群到运行模型。涵盖集群验证、控制器安装、GPU评估、提供商设置和首次部署。适用场景:“设置AI Runway”、“接入AKS集群”、“安装AI Runway”、“airunway设置”、“将模型部署到AKS”、“在AKS上进行GPU推理”、“在AKS上配置KAITO”、“在AKS上运行LLM”、“在AKS上使用vLLM”、“在AKS上设置模型服务”、“AI Runway控制器”。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights对Web应用进行插桩的指南。提供遥测模式、SDK设置和配置参考。适用场景:如何对应用进行插桩、App Insights SDK、遥测模式、什么是App Insights、Application Insights指南、插桩示例、APM最佳实践。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)为浏览器/Web应用添加检测。用于真实用户监控(RUM)——页面视图、点击、AJAX/fetch依赖项、异常、自定义事件,以及与后端OpenTelemetry追踪关联的浏览器端GenAI代理追踪。涵盖SDK加载器脚本和npm设置、框架扩展(React、React Native、Angular)、点击分析、遥测初始化器,以及从浏览器发出的代理/工具/模型跨度所遵循的OTel GenAI语义约定。
devops
azure-ai-anomalydetector-java
microsoft
使用适用于 Java 的 Azure AI 异常检测器 SDK 构建异常检测应用程序。在实现单变量/多变量异常检测、时间序列分析或 AI 驱动的监控时使用。
development
azure-ai-language-conversations-py
microsoft
使用azure-ai-language-conversations Python SDK实现对话语言理解(CLU)。当使用ConversationAnalysisClient分析对话意图和实体、构建NLP功能或将语言理解集成到应用程序中时使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用于机器学习工作区、作业、模型、数据集、计算资源和管道。 触发词:“azure-ai-ml”、“MLClient”、“工作区”、“模型注册表”、“训练作业”、“数据集”。
development