Cursor 必备 MCP 组合推荐

在 Shade 或 admin 应用中不要为颜色写 dark: Tailwind 变体——语义令牌在暗色模式下自动切换。在编辑使用 Shade 的应用的 TSX 时触发。

UI 装饰一律使用 Shade 语义令牌(bg-background、text-foreground、border-border-default、bg-surface-elevated)——绝不使用 hex、hsl() 或 bg-gray-200 之类的原始调色板工具类。在编辑使用 Shade 的应用的 TSX/CSS 时触发。

审核 Claude Code 的上下文窗口在代理、skill、MCP 服务器和规则上的消耗，识别膨胀与冗余组件，并给出按优先级排序的省 token 建议。当上下文窗口填充过快、需要找出是哪些代理、skill、MCP 服务器或规则在占用时使用。

以红-绿-重构循环用测试驱动开发。在实现任何逻辑、修复任何 bug 或改变任何行为时使用；在需要证明代码可用、收到 bug 报告或即将修改现有功能时使用。

Inventory and govern a project's test assets. Connect requirements, business rules, risks, bugs, and cross-boundary contracts to stable TEST-IDs and executable evidence in the repository's existing test registry or TESTS.md. Use for test inventory, test gap analysis, unit/integration/contract/E2E/smoke classification, bug-to-test regression protection, consumer/provider contract testing, test-necessity review, and pre-delivery evidence review.

Inventory and govern a project's test assets. Connect requirements, business rules, risks, bugs, and cross-boundary contracts to stable TEST-IDs and executable evidence in the repository's existing test registry or TESTS.md. Use for test inventory, test gap analysis, unit/integration/contract/E2E/smoke classification, bug-to-test regression protection, consumer/provider contract testing, test-necessity review, and pre-delivery evidence review.

首席安全官（CSO）模式。（gstack）

跨前端、后端、查询与数据库优化应用性能。在存在性能指标要求、怀疑性能回归、需要改善 Core Web Vitals 或加载时间、需要修复 N+1 查询模式，或剖析暴露瓶颈时使用。

把 SRS 与 Oryx 的代码任务路由到最小相关的可信代码库地图与验证指引。只要支持、开发、调试、评审或维护工作需要定位、选择、阅读、修改、测试或验证 SRS/Oryx 代码时使用。覆盖第一代 C++ 源站与边缘媒体服务器及其 State Threads 依赖、下一代 Go 服务器、浏览器推流与播放、SRS Docker 镜像工具链、SRS 测试，以及 Oryx 的 Go 后端、React 控制台、一体化运行时、打包、安装器、发布与测试。作为父级支持与发展技能的代码导航依赖；面向用户的工作流程与结果仍由父技能负责。

Fix browser/API/CLI/job/worker/webhook bugs. (gstack)

通过 Aside 浏览器从网页拉取数据——使用你真实的、已登录的会话。（gstack）

让远程 AI agent 与你的浏览器配对使用。（gstack）

启动 GStack Browser——内置侧边栏扩展、可由 AI 控制的 Chromium。

把真实 Chromium 浏览器中的 cookie 导入到无头浏览会话。（gstack）

部署后的金丝雀监控。（gstack）

设计咨询：理解你的产品、调研同类竞品，提出完整设计体系（美学、字体排印、色彩、布局、间距、动效），并生成字体与配色预览。（gstack）

把最近一次成功运行的 /scrape 流程固化为磁盘上的常驻 browser-skill。（gstack）

仅出具报告的 QA 测试。（gstack）

设计收尾：生成贴近生产质量、原生支持 Pretext 的 HTML/CSS。（gstack）

Install and operate Everything Claude Code (ECC) on the DeepSeek Harness (DSH): native skill roots (~/.dsh/skills, .agents/skills), the @deepseek-ai/dsh-hooks-claude-code bridge for command hooks, bare-insert patch mounting, generator usage, event-support limits, and update workflow. Use when setting up ECC on DeepSeek Harness, wiring ECC hooks into a DSH profile, or debugging why ECC skills or hooks do not appear in a DSH session.

Lint ECC skills for Codex/Cursor-safe frontmatter and Claude-only substitutions before a skill PR.

Lint ECC skills for Codex/Cursor-safe frontmatter and Claude-only substitutions before a skill PR.

Verify a local agent API, temporary gateway tunnel, and remote sandbox callback with a tool-free task, then restore the original app connection.

Run the ECC engineering process retroactively against an existing project slice. Audit current state, reconstruct past decisions, produce a four-axis gap register (thinking, coding practice, missing features, technology selection), then execute fixes forward through the plan-TDD-review pipeline. Use when adopting ECC mid-project, taking over an inherited codebase, or asking what was missed and what would have been built differently from day one.

Run a six-phase verification of a Claude Code session's work — build, type check, lint, tests with coverage, security grep, and diff review — then produce a PASS/FAIL verification report. Use when verifying work after completing a feature or refactor, before creating a PR, or when quality gates must pass.

Cut a milestone release: the release issue, the preparation PRs on v2/main (audit, changesets Version Packages, Python CalVer), the pure v2/main to main merge PR with its release ledger, then the GitHub Release a maintainer publishes.

当用户需要收入运营、线索生命周期管理或市场到销售的交接流程支持时使用。用户提到 RevOps、revenue operations、线索评分、线索路由、MQL、SQL、pipeline stages、deal desk、CRM 自动化、marketing-to-sales handoff、数据清洗、线索没有流转到销售、pipeline 管理、线索资格判定，或“市场何时该交接给销售”等场景时也适用。凡涉及连接市场与收入的系统和流程均可使用。冷邮件触达见 cold-email，邮件序列营销见 emails，定价决策见 pricing。

通过 gws CLI 或 Python 访问 Gmail、Calendar、Drive、Docs、Sheets。

使用 python-pptx 创建、读取与编辑 .pptx 演示文稿。

用于 SPA 路由与页面 segment：Web/桌面/移动/弹窗路由、重定向，以及 src/routes 到 features 的边界。

部署并配置 LangBot 实例——Docker / Docker Compose、Kubernetes、config.yaml 模型、Box 沙箱运行时、插件运行时，以及全局 API key。在生产或自托管环境中安装、部署、升级或配置 LangBot 时使用。触发语包括 deploy langbot、langbot docker、langbot compose、langbot kubernetes、langbot config.yaml、langbot box runtime、langbot global api key。

FastGPT 的 Docker Compose 自托管部署流程。当用户要求 AI 智能体参照 /deploy/SKILL.md 用 Docker 部署 FastGPT 时使用，包括创建空的 fastgpt 目录、以非交互模式运行安装脚本、启动 Docker Compose、监控服务健康、排查 compose 问题，并返回访问地址、root 账号凭据与后续步骤。

仅在用户明确要求审查 web/ 或 packages/dify-ui/ 下的前端代码时使用。支持待提交变更、按文件和粘贴 diff 三种审查模式。不适用于仅实现代码的请求、无审查意图的诊断、或纯后端代码。

Announcement kit for a Folo release - the X post with the highlights, its reply with the release links, a one-sentence quote line for the second account, and a 1080p promo video rendered from real app captures. The desktop and mobile release skills end with it; also use it when the user asks for promo material for a release.

Run and diagnose existing Harbor or FrontierHarness evaluations against local production LobeHub or LobeHub Cloud, on a local Docker host, SSH worker, or Runta. Use for eval infrastructure, target preflight, lh CLI injection, Harbor smoke or job runs, resume, and failure triage. Excludes authoring Harbor tasks and product acceptance.

在完成任务、实现重要功能或合并前使用，以验证工作符合要求。

Render a token-usage stat card from local Claude Code, Codex, Gemini CLI, and OpenCode session logs and commit it into the repository as a file. Use when someone wants a current usage figure in a README, or asks for a shareable card of their coding-agent activity.

Optimize LLM prompt caching hit rate to reduce API costs and improve latency. Analyzes prompt structure, identifies cacheable vs non-cacheable content, and recommends restructuring to maximize cache reuse. TRIGGER when: user says "prompt cache", "cache hit rate", "reduce LLM cost", "optimize token usage", "cheaper LLM calls", or asks about prompt caching strategies for OpenAI/Anthropic/other providers. DO NOT TRIGGER when: user just wants to optimize a single prompt's quality (use prompt-optimizer instead), or asks about general caching infrastructure.

Design or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.

Reviews staged changes and drafts a Conventional Commits-compliant commit message with the right type, scope, and length. Use before committing, or when the user asks to write, fix, or review a commit message.

Security hardening guidance for AI agent frameworks that process untrusted content, invoke tools, write workspace files, manage runtime identifiers, or handle credentials. Use when building or reviewing an agent runtime, autonomous worker, tool gateway, memory service, or multi-tenant agent deployment. Do not use for general web application security or offensive testing.

为 AI Agent 加入 x402 支付执行能力，支持按任务预算、支出管控与非托管钱包。通过 agentwallet-sdk 支持 Base，通过 OKX Payments / OKX Agent Payments Protocol 支持 X Layer。当 Agent 需要自行付费，并需要按任务预算、支出管控与非托管钱包时使用。

将 x402 支付执行添加到 AI 代理中——通过 MCP 工具实现每任务预算、支出控制和非托管钱包。当代理需要为 API、服务或其他代理付费时使用。

让 AI 代理执行 x402 支付：按任务限额、支出控制、非托管钱包。通过 agentwallet-sdk 支持 Base，通过 OKX Payments / OKX 代理支付协议支持 X Layer。

Create and configure hookify rules — markdown files with YAML frontmatter that match bash, file, prompt, or stop events by regex or conditions and show warn/block messages to the agent. Use when creating a hookify rule, writing hook rule syntax, configuring hookify, or adding pattern guardrails such as blocking dangerous commands, .env edits, or debug code.

Claude API / Anthropic SDK 参考——模型 id、定价、参数、流式输出、工具调用、MCP、agent、缓存、token 计数、模型迁移。触发条件：提示中出现任何 Claude/Anthropic 相关名称（Claude、Anthropic、Fable、Opus、Sonnet、Haiku 等）；用户询问 LLM 相关问题（定价/选型/限额/缓存）时不要凭记忆回答；或任务属于 LLM 类且未指明供应商。仅当明确针对其他供应商（OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama）时跳过。

Diagnose GKE Cloud TPU training throughput drops and step-time regressions (15%+ TPU duty-cycle drop) using ML Diagnostics Workload Monitoring (`gcloud alpha mldiagnostics monitored-events` / `hypercomputecluster.googleapis.com/v1alpha`) and 1-minute Cloud Monitoring system metrics (`kubernetes.io/node/accelerator/*`). Distinguishes hardware and network fabric throttling from workload resource bottlenecks (HBM capacity, host memory, or host CPU saturation). Use when TPU training throughput or duty cycle drops without crashing pods, when `PERFORMANCE_DEGRADATION` monitored events fire, or when triaging slow multi-slice training steps. Don't use for complete multi-slice XLA execution stalls with `HANG_DETECTED` logs (use gke-ai-troubleshooting-tpu-mxla-hang) or pod eviction/interruption restarts (use gke-ai-troubleshooting-jobset-interruption).

Draw diagrams for the Next.js docs in the style of the ones already published there: the light/dark PNGs an mdx references with `<Image srcLight="/docs/light/<name>.png" srcDark="/docs/dark/<name>.png">`. Use when a guide or API reference needs a new diagram or an existing one needs a change. Hand-written SVG from a primitives library, rendered by headless Chrome at 2x. No AI image generation.