程序员努力
AI导航AI编程实战PromptMCP市场Skills市场

程序员努力

AI 编程资源库 - 聚合 AI 编程教程、提示词、MCP 与 Skills 资源

内容栏目

  • AI导航
  • AI编程实战
  • Prompt
  • MCP市场
  • Skills市场
  • 全站搜索

教程分类

  • 进阶技巧
  • 入门指南
  • 最佳实践
  • Claude Code
  • Cline教程
  • GitHub Copilot
  • Cursor教程
  • 提示词工程

关于

  • 关于我们
  • GitHub

© 2026 程序员努力. All rights reserved.

冀ICP备15004808号-7

首页教程PromptMCPSkills
← 返回Skills市场
R

Review-Gate

Review-Gate高频

by LakshmanTurlapati

综合评分详见右侧面板

Review-Gate V2 is a powerful rule for the Cursor IDE that helps you get up to 5x more value from your monthly requests. It creates an interactive loop where the AI waits for your …

效率

安装此 Skill

一键安装(GitHub 托管)
git clone https://github.com/LakshmanTurlapati/Review-Gate ~/.claude/skills/review-gate

简介

Review-Gate 是一个针对 Cursor IDE 的规则工具,通过 MCP 集成创建交互式循环,让 AI 在单个请求内等待用户的后续指令(文本、语音或图像上传),从而将每月请求的利用率提升至最高 5 倍。它支持本地 Whisper 语音转文字、图像上下文上传以及美观的弹窗界面,帮助用户在同一请求生命周期内完成深度迭代工作,避免频繁消耗请求配额。

SKILL.md 预览

3OtOp7R.th.png

Review Gate for Cursor IDE

Version License Platform MCP Whisper Stars Forks

Review Gate V2 Interface

Cursor would often drop the mic way too early! I'd give it a complex task, it'd use maybe 5 of its ~25 available tool calls for that single "main request," then call it a day. Not only was that untapped AI power for that single thought, but making small follow-up tweaks meant starting a new request. Doing that too often, and my precious ~500 monthly requests (you know the ones!) would burn up much faster than I liked.

Presenting: The Review Gate – The "Turn Your 500 Cursor Requests into 2500!" Rule! (Okay, maybe not always a perfect 5x, but you get the idea!)

I evolved this system for our beloved Cursor IDE to transform my (and your!) AI from a quick sprinter into an endurance marathon runner for complex ideas, all within the lifecycle of a single main request. Now with V2's voice commands, image uploads, and beautiful popup interface! I've basically told Cursor: "Hold up, we're not done with this request until I say we're done."

Architecture

graph TD
    A[User Gives Complex Task] --> B[Cursor Agent Processes Task]
    B --> C[Review Gate Rule Activates]
    C --> D[Popup Interface Appears]
    D --> E[Multi-Modal User Input]
    E --> E1[Text Commands]
    E --> E2[Voice Input]
    E --> E3[Image Upload]
    E1 --> F[Agent Continues with More Actions]
    E2 --> F
    E3 --> F
    F --> G{Task Complete?}
    G -->|No| D
    G -->|Yes| H[Single Request Delivers 5x Value]
    

Features Comparison

| Feature | V1 (Terminal) | V2 (MCP Integration) | |---------|--------------|---------------------| | Interface | Terminal-based script | Beautiful popup interface | | Input Methods | Text only | Text + Voice + Images | | Integration | Python script | Native MCP tool | | Platform Support | macOS/Linux | macOS + Windows + Linux | | Speech-to-Text | No | Yes - Local Whisper AI | | Image Upload | No | Yes - Multi-format support | | Real-time Status | No | Yes - MCP status indicator | | Installation | Manual rule setup | One-click installer |

Key Features (What V2 Packs In)

  • Voice-Activated AI Control: Speak your sub-prompts directly! Click the mic, speak naturally, and watch your words transcribe automatically using local Faster-Whisper AI.
  • Visual Context Sharing: Upload images, screenshots, diagrams, or mockups directly in the popup. The AI sees everything you share.
  • Beautiful Popup Interface: Professional orange-glow design that fits perfectly in Cursor with real-time MCP status indicators.
  • AI On MY Leash: Makes the Cursor Agent wait for my (and your!) "go-ahead" via an interactive popup before it truly signs off on an initial request.
  • Multiply Your Request Power: Make one main request do the work of many! Instead of 5 new prompts (and 5 dings on your ~500 request counter!), use the Review Gate for 5 (or more!) iterative sub-prompts within that single request's lifecycle and tool call budget.
  • Unlock Full Tool Call Potential: I designed this to help us guide the AI to use more of its ~25 available tool calls for a single complex idea through those sub-prompts.
  • MCP Integration Magic: Built on the Model Context Protocol for seamless Cursor integration. The popup automatically appears when needed.
  • Cross-Platform Speech: Whisper speech-to-text works flawlessly on macOS and is implemented for Windows and Linux.

How It Works (V2 Evolution)

  1. You (or I): Give Cursor a task (this counts as 1 main request towards your ~500).
  2. Cursor AI: Does its main job (coding, analysis, maybe a few tool calls from the ~25 for this request).
  3. Review Gate V2 Kicks In (The Magic Part I Evolved!):
    • AI calls the review_gate_chat MCP tool automatically
    • Beautiful popup appears in Cursor with multi-modal input options
    • AI announces it's waiting for your input in the popup
  4. You (in the popup):
    • Type quick follow-ups (e.g., "Now add docstrings to all new functions.")
    • Speak your commands using the microphone (automatic transcription)
    • Upload images for visual context (screenshots, mockups, diagrams)
    • Or type TASK_COMPLETE when you're satisfied
  5. Cursor AI (powered by MCP integration): Reads your popup input (text, speech, images), acts on it (more coding, more tool calls from the original budget!), responds in the main chat, then opens the popup again for your next input.
  6. Loop! This continues, deepening the work on your original request, until you type TASK_COMPLETE in the popup.

Installation

For the full V2 experience, start with V2/INSTALLATION.md. It documents the canonical shipped V2 files, including V2/review-gate-v2-2.7.3.vsix, V2/ReviewGateV2.mdc, the manifest-backed packaging helper, and the merge-safe MCP setup.

V2 Installation (Recommended - MCP Integration)

Two simple steps to supercharge your Cursor workflow:

Step 1: One-Click Technical Setup

macOS Installation

# Clone repository and navigate to V2 subdirectory
git clone https://github.com/LakshmanTurlapati/Review-Gate.git
cd Review-Gate/V2

# Run the magical one-click installer
./install.sh

Windows Installation

# Clone repository and navigate to V2 subdirectory
git clone https://github.com/LakshmanTurlapati/Review-Gate.git
cd Review-Gate/V2

# Run PowerShell installer (may need admin privileges)
./install.ps1

The installer automatically handles:

  • Dependencies: Package managers (Homebrew/Chocolatey), SoX (for speech), Python packages
  • MCP Server: Global installation with proper configuration
  • Extension: Cursor extension for the popup interface
  • Configuration: MCP integration setup that preserves existing MCP configuration by merging only the review-gate-v2 entry

Maintainer Packaging Contract

If you need to validate or rebuild the shipped artifact, the repo now treats V2/review-gate-v2-2.7.3.vsix as the only canonical release file:

# From the repo root
python3 scripts/package_review_gate_vsix.py --check

# From V2/cursor-extension
npm run package

Both commands read V2/release-manifest.json, keep V2/ReviewGateV2.mdc as the supported rule file, and target the same canonical VSIX path under V2/.

GitHub Actions Build

The repo also includes .github/workflows/build-vsix.yml for CI packaging. On pushes to main, pull requests, or manual dispatch, it:

  1. Validates V2/release-manifest.json
  2. Runs python3 scripts/run_review_gate_regression_checks.py --suite release-surface
  3. Builds the canonical VSIX with npm run package
  4. Uploads V2/review-gate-v2-2.7.3.vsix as a workflow artifact

This workflow is intentionally build-only for now. It does not publish a GitHub Release, and it does not change the installation steps above.

Alternative: Manual Extension Installation

If the automatic installer doesn't install the extension properly, you can install it manually:

  1. Download Extension: Use V2/review-gate-v2-2.7.3.vsix, the canonical shipped artifact
  2. Open Cursor: Launch Cursor IDE
  3. Extensions Panel: Press Cmd/Ctrl+Shift+X to open Extensions
  4. Install from VSIX: Click the ... menu → "Install from VSIX..."
  5. Select File: Choose the downloaded .vsix file
  6. Restart Cursor: Restart Cursor completely for the extension to activate

Step 2: Copy the V2 Rule to Cursor

CRITICAL STEP: For the Review Gate V2 to work, you need to copy the V2 rule to your Cursor settings:

  1. Open the Rule File: Copy the entire contents of V2/ReviewGateV2.mdc
  2. Cursor Settings: Open your Cursor IDE → Settings (Cmd/Ctrl + ,)
  3. Find Rules Section: Look for "Rules" or "AI Rules" in the settings
  4. Paste & Save: Paste the entire V2 rule content and save
  5. Restart Cursor: Restart Cursor completely for the rule to take effect

If you ever edit ~/.cursor/mcp.json manually, preserve existing MCP server entries and merge the review-gate-v2 configuration instead of replacing the whole file.

V1 Installation (Original Terminal-Based)

For those who prefer the original terminal-based approach:

  1. Copy THE Rule: You'll need the "Review Gate" rule text from the V1 documentation.
  2. Cursor Settings: Open your Cursor IDE settings.
  3. Global AI Rules: Find the section for "Rules".
  4. Paste & Save: Paste the entire rule. Save.
  5. Boom! The original Review Gate is now armed to make every main request count for way more.

Testing Your Installation

After both steps are complete:

  1. Manual Popup Test: Press Cmd+Shift+R in Cursor to open the popup manually
  2. Agent Integration Test: Ask Cursor: "Use the review_gate_chat tool to get my feedback"
  3. Speech Test: Click the microphone → speak clearly → verify automatic transcription
  4. Image Test: Click the camera icon → upload an image → send with text
  5. Full Workflow Test: Give Cursor a complex task and watch the Review Gate V2 popup appear automatically

What You Get (V2 Feature Summary)

Voice-to-Text Magic

  • Click microphone → speak naturally → automatic transcription
  • Local Faster-Whisper AI processing (no cloud, no privacy concerns)
  • Professional visual feedback: mic → red stop button → orange spinner → text injection

Image Upload Power

  • Support for PNG, JPG, JPEG, GIF, BMP, WebP formats
  • Drag & drop or click to upload
  • Images included in MCP responses so the AI can see your visual context
  • Perfect for sharing screenshots, mockups, error dialogs, or architectural diagrams

Beautiful Interface

  • Clean popup with orange glow design that matches Cursor's aesthetic
  • Horizontally aligned controls for professional appearance
  • Real-time MCP status indicator
  • Smooth animations and responsive feedback

Seamless MCP Integration

  • Works automatically with Cursor Agent tool calls
  • 5-minute timeout for thoughtful responses
  • Global installation works across all your Cursor projects
  • File-based communication protocol for reliability

Play Smart (My Tips & The "Why")

  • Why I evolved this hack: To stop Cursor from ending too soon when I have iterative follow-ups for the same original thought process, but now with the power of voice commands and visual context sharing.
  • Voice Commands Work Best When: You speak clearly and concisely. The local Whisper AI is quite good but prefers natural, well-paced speech.
  • Image Context is Gold: Upload screenshots of errors, mockups of what you want built, or diagrams of architecture. The AI can see and understand visual context.
  • Platform Notes: Speech-to-text is rock-solid on macOS. Windows and Linux implementations exist and are improving.
  • Be Clear in All Inputs: Whether typing, speaking, or sharing images, clear and direct communication in the popup works best.
  • TASK_COMPLETE is Your Exit: Don't forget to type this in the popup to let the AI finally rest (and free up that main request slot).

Troubleshooting

V2 Troubleshooting

# Check if MCP server is running
tail -f /tmp/review_gate_v2.log

# Test speech recording capability
sox --version
sox -d -r 16000 -c 1 test.wav trim 0 3 && rm test.wav

# Check extension logs in Cursor
# Press F12 → Console tab for browser logs

# Verify MCP configuration
cat ~/.cursor/mcp.json

# Test the extension manually
# Press Cmd+Shift+R in Cursor

Heads Up! (My Friendly Warnings)

  • EXPERIMENTAL EVOLUTION! This is V2 of my power-user move. It works because we're very cleverly instructing the AI with MCP integration.
  • MCP SERVER RUNS LOCALLY: The rule uses a local MCP server that integrates with Cursor. The installer sets this up automatically.
  • SPEECH PROCESSING IS LOCAL: Your voice is processed locally using Faster-Whisper AI - nothing goes to the cloud.
  • PLATFORM COMPATIBILITY:
    • macOS: Primary development path with automated coverage; native installer and live Cursor validation remain tracked follow-up checks
    • Windows: Implemented and supported, with native shell smoke still environment-dependent
    • Linux: Supported path, but less battle-tested than macOS
  • PYTHON & SOX NEEDED: The installer handles these, but your system needs to support Python 3 and SoX for speech functionality.
  • CURSOR UPDATES MIGHT CHANGE THINGS: Future Cursor versions could affect how this rule behaves. What works today might need tweaks tomorrow!
  • REMEMBER THE RULE: The MCP server is just the engine - you MUST copy the V2 rule to your Cursor settings for automatic activation!

Why V2? You Asked For It!

After the original Review Gate gained 100+ forks, 100+ stars, 100,000+ impressions, and countless requests to evolve it further - you asked for it, I delivered! The community response was incredible, and the demand for voice commands, visual context, and a more sophisticated interface was overwhelming. V2 is my answer to every single feature request and improvement suggestion.

About Me & This Evolution

This "Review Gate V2" represents the evolution of my original terminal-based rule into a full-featured, multi-modal interaction system. It was born from my own desire to truly partner with Cursor's AI using not just text, but voice commands and visual context. My goal remains the same: to make every interaction as deep and complete as possible and ensure every available tool call for a big idea gets its chance to shine, making each of those ~500 requests count like gold!

The V2 system leverages the Model Context Protocol (MCP) to create a seamless bridge between Cursor's AI and a rich, interactive popup interface. Whether you're speaking commands, sharing screenshots, or typing follow-ups, it's all designed to keep you in the flow while maximizing the value of each Cursor request.

To connect with me or learn more about my work, visit: www.audienclature.com


Happy (and supercharged) coding with Cursor! May your AI always await your final command, your voice be clearly transcribed, your images perfectly understood, and your monthly requests feel like they last forever!

统计信息

高频

使用频率(自动计算)

159

Fork

JavaScript

语言

2026/4/2

更新

综合评分

2.9
社区认可
4
使用热度
3
文档完整
3
功能丰富
1
维护活跃
2

基于社区认可、使用热度、文档完整度、功能丰富度、维护活跃度自动计算

标签

CursorMCP效率提升语音控制图像上传交互式开发

链接

GitHub

相关Skills

a

agent-rules

agent-rules

Rules and Knowledge to work better with agents such as Claude Code or Cursor

效率中频
a

awesome-cursorrules

awesome-cursorrules

📄 Configuration files that enhance Cursor AI editor experience with custom rules and behaviors

效率中频
a

architecture-copilot

architecture-copilot

🧭 给 Claude Code / Cursor / Codex 用的『架构副驾』skill —— 开新项目时用持续深度提问引导你在写代码前想清楚架构(产出架构图 / ADR / 演进路线)。知识源自 awesome-architecture。中英双语。

效率中频
g

graphify

graphify

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local determi…

效率中频