菜单
Configure MCP (Model Context Protocol) servers for Claude or Cursor. Use when the user runs /setup, wants to add Atlassian, Datadog, or Playwright MCP, or set up API keys for MCP connections.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/micaelmalta/agentic --skill setup命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | setup |
| description | Configure MCP (Model Context Protocol) servers for Claude or Cursor. Use when the user runs /setup, wants to add Atlassian, Datadog, or Playwright MCP, or set up API keys for MCP connections. |
| triggers | ["/setup","setup MCP","configure Atlassian MCP","configure Datadog MCP","configure Playwright MCP","add MCP servers","connect to Atlassian","connect to Datadog","setup Playwright"] |
Configure Atlassian (two servers: atlassian, atlassian-tech), Datadog, and Playwright MCP servers for Claude or Cursor so the agent can use Jira, Confluence, Datadog (logs, metrics, monitors), and browser automation for UI testing during workflows.
For full MCP management (add, remove, list, auth, scopes), use the official docs:
| Client | Docs |
|---|---|
| Claude (Claude Code) | Connect Claude Code to tools via MCP — claude mcp add/list/remove, HTTP/stdio/SSE, scopes (local/project/user), OAuth via /mcp. |
| Cursor | Model Context Protocol (MCP) — mcp.json (project: .cursor/mcp.json, global: ~/.cursor/mcp.json), stdio/HTTP/SSE, config interpolation, OAuth. |
Refer to these when the user asks how to manage MCP in general, add other servers, or troubleshoot.
/init)Before MCP configuration, run /init for PARA so the project has a consistent workflow context:
context/ with context.md, plans/, summaries/, archives/, data/, servers/ (create any missing dirs and context/context.md). Note: context/ is git-ignored so .gitkeep files are not needed.CLAUDE.md exists (create a minimal one if missing).If the structure already exists, skip or only add missing pieces. Then proceed with MCP setup below.
Ask the user which client(s) to configure. Use the official config locations:
| Option | Config location | Notes |
|---|---|---|
| Cursor | ~/.cursor/mcp.json | Cursor MCP docs |
| Claude (Claude Code CLI) | ~/.claude.json | Claude Code MCP docs; use claude mcp add or merge into this file |
| Claude Desktop (app) | ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) | Set CLAUDE_DESKTOP=1 when running the setup script |
| Both | Cursor + Claude Code paths | Write to both ~/.cursor/mcp.json and ~/.claude.json |
If the user doesn’t specify, offer Cursor and Claude and default to both (Cursor + Claude Code) unless they say otherwise.
Prompt for credentials. Prefer environment variables for security; accept one-time input if the user prefers.
Datadog (required for Datadog MCP):
Atlassian (optional):
Playwright (no keys required):
--headless flag for headless mode (useful in CI)--browser flag to specify browser (chrome, firefox, webkit, msedge)If the user already has Cursor MCP configured (e.g. OAuth for Atlassian), keep existing mcpServers and only add or update atlassian, datadog, and playwright entries.
command: nodeargs: ["<YOUR_MCP_DATADOG_PATH>/src/index.js"]
If the user has a different path (e.g. MCP_DATADOG_PATH), use that in args.url: https://mcp.atlassian.com/v1/mcp, type: httpATLASSIAN_TECH_URL in env when running the setup script if the tech endpoint differs
Add env (e.g. ATLASSIAN_API_TOKEN, ATLASSIAN_EMAIL) only if the user provides them and the client supports env-based auth for HTTP MCP.command: npxargs: ["@playwright/mcp@latest"]
Optional args: --headless, --browser <browser>, --viewport-size <WxH>, --config <path>Cursor (~/.cursor/mcp.json):
mcpServers exists.mcpServers.atlassian: { "url": "https://mcp.atlassian.com/v1/mcp", "type": "http" } (and optional env if provided).mcpServers["atlassian-tech"]: same structure; URL from ATLASSIAN_TECH_URL if set.mcpServers.datadog: { "type": "stdio", "command": "node", "args": ["<path-to-mcp_datadog/src/index.js>"], "env": { "DATADOG_API_KEY": "<user-key>", "DATADOG_APP_KEY": "<user-app-key>" } }.mcpServers.playwright: { "command": "npx", "args": ["@playwright/mcp@latest"] } (add --headless to args for CI environments).atlas, cursor-ide-browser).Claude Code (~/.claude.json):
mcpServers: add or update atlassian, datadog, and playwright. This is the correct config for the Claude Code CLI (claude). Preserve existing mcpServers and other top-level keys.Claude Desktop (optional; ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
mcpServers. Use only when the user targets the Claude Desktop app; the setup script uses this path when CLAUDE_DESKTOP=1 is set.The skill includes skills/setup/setup_mcp.sh and skills/setup/setup_mcp.js, which use the official config locations:
DATADOG_API_KEY, DATADOG_APP_KEY, optional MCP_DATADOG_PATH, optional TARGET (cursor|claude|both), optional CLAUDE_DESKTOP=1 for Claude Desktop app.~/.cursor/mcp.json (per Cursor MCP docs).~/.claude.json (Claude Code CLI) by default, or into ~/Library/Application Support/Claude/claude_desktop_config.json when CLAUDE_DESKTOP=1 (per Claude Code MCP docs).mcpServers; never log keys (keys come from env only).Usage:
export DATADOG_API_KEY="your-api-key"
export DATADOG_APP_KEY="your-app-key"
export TARGET="both" # or "cursor" or "claude"
# Optional: export CLAUDE_DESKTOP=1 # use Claude Desktop app config instead of Claude Code
./skills/setup/setup_mcp.sh
Run from the project root so the path to the script is correct (e.g. ./skills/setup/setup_mcp.sh or bash skills/setup/setup_mcp.sh).
Quick reference:
atlassian, atlassian-tech) for Jira/Confluencenpx @playwright/mcp@latest for UI testing, optional --headless for CIFor complete configuration examples: See reference/CONFIG_EXAMPLES.md, which covers:
mcpServers entries)After setup, the workflow, ci-cd, debugging, documentation, and testing skills can use these MCPs (see those skills for when to call Atlassian/Datadog/Playwright tools). Use atlassian or atlassian-tech tools as appropriate (e.g. atlassian-tech for tech-specific Jira/Confluence use). Use playwright for UI testing and browser automation.
/init) before MCP configuration.mcpServers entries preserved (not overwritten).| Situation | Skill to invoke | How |
|---|---|---|
| Atlassian MCP configured → use in workflow | workflow skill | Read skills/workflow/SKILL.md |
| Datadog MCP configured → use for monitoring | ci-cd / debugging skill | Read skills/ci-cd/SKILL.md or skills/debugging/SKILL.md |
| Playwright MCP configured → use for testing | testing skill | Read skills/testing/SKILL.md |
| Need to build a custom MCP server | mcp-builder skill | Read skills/mcp-builder/SKILL.md |
| MCP config needs documentation | documentation skill | Read skills/documentation/SKILL.md |
Complete development workflow from planning to GitHub PR using PARA methodology and RLM for large codebases. Use when implementing a feature, fixing a bug, refactoring, or taking work from plan to GitHub PR.
Analyze and improve performance: profile, find bottlenecks, optimize, and instrument code with observability for diagnosing performance issues (profiling, bottleneck tracing). Use when the user asks about performance, slow code, bottlenecks, profiling, optimization, or adding performance-specific observability.
Design and implement tests (unit, integration, e2e, accessibility, i18n), improve coverage, and run test suites. Use when the user asks to add tests, write tests, improve coverage, run tests, test accessibility, test translations, or implement test strategy.