367 boosters for "llm" — open source, verified from GitHub, ready to install
用户学完一个东西想验真伪，或隐约觉得"好像懂了但不踏实"。也是"重输入轻输出"的一次强制输出。 请他用自己的话、把你当外行，把概念讲一遍。别让他背定义——要他解释、打比方。 专挑他含糊带过、用术语糊弄、跳过的环节追问："为什么？""那这个是怎么来的？""举个例子？"命中他答不上来或开始绕的地方。
用户在纠结"要不要学 X / 学到什么程度 / 精力往哪放"。这是"广度优先、兴趣队列过长"倾向的刹车。 逼问一句：你要解决的具体问题是什么？ 没有具体问题、纯"感觉该学 / 别人都在学"→ 直接进"以后再学"队列，不占当下精力。理解知识的作用，重于知识本身。 这知识多久会贬值？（技术栈 / 工具往往 6–12 个月就明显更新）相对有限的时间值不值？贬值快 + 可外包给 AI / 随时查 → 只需"知道它存在、管什么"，不必真学。
用户在学 / 接触一个新概念 X（新技术、新算法、新理论、新领域……），尤其觉得"陌生 / 有点难"的时候。难，往往不是智商问题，是它相对用户还存在"没接上的旧知识"。 用一两句话说清 X 到底在干什么——它的核心机制 / 结构是什么。剥掉术语外壳，留下"它本质是一个 "。只有先拿到结构，才能去匹配用户学过的东西。 拿不准就直接问「你学过 吗？」，绝不从正在讲的材料 / 文章作者背景推断用户会什么。
梅花易數（梅花心易）專業占卜系統。相傳為宋代邵雍（邵康節）所創。 占卜目的不只告知吉凶，而是指引如何做出最佳改變。 每次解卦必須提供策略建議，告訴用戶下一步該如何行動。
Native Claude Agent SDK integration for AgentPool that enables file operations, terminal access, and advanced code editing with lower latency and tighter integration than bridge alternatives. Developers building AI agents on Claude Code or Claude Desktop benefit from direct SDK access and streamlined tooling.
A MCP Server that integrates Langfuse prompt management capabilities into Claude Desktop and Claude Code, enabling developers to access and manage prompts directly within their AI workflows.
Recall MCP Server enables persistent, cross-session memory for Claude and AI agents through Redis/Valkey or managed cloud hosting, allowing AI systems to retain and retrieve context across conversations. Developers building Claude applications, multi-turn agents, and AI systems requiring long-term context management benefit most from this solution.
"name": "llm-wiki-compiler", "name": "llm-wiki-compiler", "source": "./plugin",
"name": "claude-prompts", "description": "Hot-reloadable versioned prompts with easy tools for prompt engineering, chain workflows, quality gates. Symbolic syntax: >>prompt --> >>chain @framework :: 'gate'", "homepage": "https://github.com/minipuft/claude-prompts",
Context-Keeper is a Cursor rules booster that implements an intelligent memory and context management system with a structured coding execution framework emphasizing thinking before coding, user confirmation, and iterative quality assurance. It benefits developers who need better workflow discipline and context awareness when using Cursor AI.
<summary >🌐 Language</summary> <div align="center"> <a href="https://openaitx.github.io/view.html?user=wshobson&project=agents&lang=en">English</a>
OK Computer is a self-hosted AI agent system prompt that configures Claude as a general-purpose agent with specialized rules for data analysis, slide creation, and web deployment. Developers building agent workflows on Claude-compatible platforms benefit from its structured tool integration and file management policies.
去看 .kilocode\rules\rules.md
Sub-Agents enables parent agents to delegate specialized tasks to child agents with independent prompts, tools, and providers, allowing developers to build modular, hierarchical AI systems that solve complex multi-step problems.
A concurrent sub-agent executor that transparently parallelizes independent tool calls in agentic workflows, enabling faster multi-step operations without explicit LLM coordination. Developers building multi-agent systems in Claude Code or Claude Desktop benefit from improved performance and simplified agent orchestration.
"name": "@cyanheads/mcp-ts-core", "version": "0.2.11", "mcpName": "io.github.cyanheads/mcp-ts-core",
You are "TunaCode", a Staff-level software developer agent inside the user's terminal. You are not a chatbot. You are an operational agent: you search, read, write, and execute code. Think step by step. Be direct, neutral, and concise.
This MCP server bridges n8n workflow automation with Claude by providing programmatic access to workflows, executions, and credentials through the Model Context Protocol. It's valuable for developers and automation engineers who want to integrate n8n capabilities into AI-powered applications and Claude instances.
"name": "matryoshka-rlm", "version": "0.2.17", "description": "Recursive Language Model - Process documents larger than LLM context windows",
This booster provides best practice guidelines for Docker, npx, uv, and package publishing workflows, intended to standardize development processes in Windsurf environments. It benefits developers seeking consistent build and deployment practices across multiple tool ecosystems.
Native SQL extensions for GoldenMatch ().
Dinox MCP Server enables Claude to understand images through advanced object detection, localization, and captioning by connecting to the DINO-X vision model. Developers building multimodal AI applications benefit from seamless integration of real-world visual perception into LLM workflows.
"name": "firefox-devtools-mcp", "description": "Model Context Protocol (MCP) server for Firefox DevTools automation", "author": "Mozilla",
Systematic REST API development with error handling, validation, and consistent response formats. This Skill enforces: Apply when building API routes, handling errors, or designing responses.
