thumbgate

Convert thumbs-down feedback into deterministic pre-action blocks for MCP tool calls.

25|7|Updated Mar 3, 2026
One-click install
npx skills add https://github.com/IgorGanapolsky/ThumbGate --skill thumbgate
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: thumbgate
Source: https://github.com/IgorGanapolsky/ThumbGate/tree/main
Command: npx skills add https://github.com/IgorGanapolsky/ThumbGate --skill thumbgate

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

ThumbGate prevents AI coding agents from repeating known, high-risk mistakes by converting thumbs-up/down feedback into deterministic pre-action blocks before tool calls execute.

Core Features & Use Cases

  • Thumbs-down to Prevention Rules: Captures failure context and auto-promotes repeated failures into prevention rules.
  • Pre-Action Tool Gating (PreToolUse): Intercepts tool calls via MCP hooks and blocks known-bad actions before they run.
  • Context Packs & Retrieval: Provides bounded relevant lesson history for the current task to improve consistency without requiring model fine-tuning.

Quick Start

Install ThumbGate for your agent by running: npx thumbgate init, then give your agent a thumbs-down on a repeated failure so the next matching action is blocked.

Frequently Asked Questions about thumbgate

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent AI coding agents from repeating the same tool use mistakes?

You can prevent AI coding agents from repeating tool use mistakes by capturing thumbs-down feedback and enforcing deterministic blocking via PreToolUse hooks before execution. ThumbGate intercepts MCP tool calls and blocks known-bad actions automatically.

Can I use PreToolUse hooks to block unsafe tool execution in Cursor or Cline?

Yes, you can use PreToolUse hooks to block unsafe tool execution in MCP-compatible agent workflows like Cursor, Cline, Codex, Gemini CLI, Amp, and OpenCode. ThumbGate intercepts and gates tool calls before they run.

How do I turn thumbs-down feedback into prevention rules for AI agents?

To turn thumbs-down feedback into prevention rules, run npx thumbgate init. The system captures failure context and auto-promotes repeated failures into deterministic rules that block future matching actions.

What is the best way to reduce wasted tokens from repeated AI agent failures?

The best way to reduce wasted tokens from repeated AI agent failures is to implement pre-action tool gating. ThumbGate applies bounded context packs and retrieval to provide relevant lesson history without requiring model fine-tuning.

Do I need to fine-tune my model to stop AI agents from making high-risk mistakes?

No, you do not need to fine-tune your model to stop AI agents from making high-risk mistakes. ThumbGate provides bounded relevant lesson history for the current task to improve consistency without model fine-tuning.

When should I not use a pre-action gating approach for coding workflows?

You should not use a pre-action gating approach for coding workflows when your agent operates outside MCP-compatible environments, as ThumbGate relies on MCP hooks and PreToolUse interception to enforce blocking rules effectively.