73 boosters for "learning" — open source, verified from GitHub, ready to install
用户要动手做 / 研究一个东西，或想把某个已有产出改得更好。这是"重输入、轻输出"短板的解药——逼用户从输入切到输出。 别追求完美，先有一个能跑 / 能看的最小版本。卡在"还没准备好"就是没进改良主义。 针对缺陷提一个改良策略（视为假说，可对可错），动手改，看效果。错了也有用——错误暴露后，下次自动规避这个方向。
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。
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。
基于 Benjamin Bloom「2 Sigma Problem」研究（1984）的一对一 AI 导师系统。每个课题是一个独立文件夹，通过自适应生成的课程文档 + 用户反馈循环模拟一对一苏格拉底式导师，把学习效果推向 +2σ。学习的主要载体是文档，对话只是辅助确认状态。 所有回复、解释、提问、文档一律使用中文。 触发本 skill 后，以下守则在整个学习交互全程生效——违反字面就是违反精神：
用户说"想学 / 理解 / 搞懂 / 讲讲一个概念 X"时——这是默认入口，一次跑完五视角，用户再选深入哪个。 抓住 X 的本质结构（剥术语），按三猜想给 🎁其实已学过 / 🔗结构同构（字段级对应表）/ 🧩可用已有知识解释，点出元知识。先激发信心，再谈深入。 定位"既定问题"（学 X 解决什么）、现有知识够不够、X 的贬值速度与 ROI，给"够用就停 / 只学最小那块 / 值得深挖"的深度边界。不是劝退，是防止一上来过度钻。
This skill transforms Claude into an interactive assistant for capturing and analyzing Claude Code's API communications using mitmproxy, a free, open-source HTTPS proxy tool. The skill supports three equally important use cases: 1. Learning & Exploration - Understanding Claude Code's internal workin
Code A2Z is a collaborative blogging platform built as a monorepo with separate client and server applications. Contributors can create, manage, and share blog posts about their projects with markdown support, customizable templates, and role-based access control. Each feature module follows this pa
"name": "compound-knowledge-plugin", "url": "https://every.to" "description": "Workflows for knowledge work that compounds over time",
Memelord is a persistent memory system for AI coding agents that uses vector search and reinforcement learning to help agents learn from past interactions. It's useful for developers building sophisticated coding assistants that need to retain and leverage historical context.
DyNote 是面向 Codex 这类 Agent 的抖音学习工具，不是一次性摘要器。核心原则是“数据资产先行，学习笔记后置”：先把字幕/转写、评论、元数据和 AI 快读沉淀为可复用资产，再按用户需求生成可追溯的学习笔记、总结和写作材料。默认目标不是字幕工程文件，而是先落一份原始数据包：、、、、、、，并用 归档可复用资产。默认把独立字幕轨或本地自动语音识别转写当作事实主干；当转写密度低、任务需要画面理解或用户只要快速筛选时，再用已登录抖音网页版的“问AI / 识别画面”补充，豆包只作为抖音 AI 不可用时的备用快读或待核验假设。 联网或登录态操作必须先使用 。不要读取、复制或打印 Cooki
This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, and model persistence. Should be invoked for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.
Automatically captures session learnings, decisions, and context into markdown files to help future agents quickly understand prior work and decisions. Developers and teams benefit from persistent knowledge transfer across work sessions.
"name": "prism-mcp-server", "mcpName": "io.github.dcostenco/prism-mcp", "description": "The Mind Palace for AI Agents — persistent memory (SQLite/Supabase), behavioral learning & IDE rules sync, multimodal VLM image captioning, pluggable LLM providers (OpenAI/Anthropic/Gemini/Ollama), OpenTelemetry
"name": "coworkpowers", "description": "Knowledge work superpowers that compound over time. Research, execute, review, and capture learnings to make each task easier than the last.", "name": "Nabeel Hyatt",
A Cursor rules file attempting to provide a memory engine interface with transformer-based architecture, but lacks concrete implementation and practical integration guidance for actual use.
"name": "claude-tutor", "description": "Interactive learning companion — creates personalized learning plans, quizzes with adaptive difficulty, and tracks progress across sessions", "keywords": ["learning", "education", "quiz", "study", "tutor"],
You are the solana-guide, an educational specialist for Solana blockchain development. You teach understanding, not memorization, through progressive learning and practical examples. 1. Teach Understanding, Not Memorization 2. Progressive Complexity
"name": "equilateral-agents", "description": "EquilateralAgents - 22 self-learning AI agents with community standards contribution. Features agent memory, pattern recognition, and workflow optimization for security, quality, deployment, and compliance.", "name": "HappyHippo.ai",
"description": "Production-ready Claude Code configuration with role-based workflows (PM→Lead→Designer→Dev→QA), safety hooks, 44 commands, 19 skills, 8 agents, 43 rules, 23 hook scripts, auto-learning pipeline, hook profiles, and multi-language coding standards", "name": "xiaobei930", "url": "https:
A PRD specialist agent that generates comprehensive product requirements documents by analyzing project patterns, assessing feature complexity, and researching best practices across your codebase. Ideal for product managers, tech leads, and developers who need structured, pattern-aware PRDs quickly.
Educational guide for Solana development concepts. Teaches programming patterns, explains code, creates tutorials, and designs learning paths for developers at all levels. Use when: Explaining Solana concepts, creating tutorials, designing learning paths, or helping developers understand complex blockchain code and patterns.
"name": "learnship", "description": "Agentic engineering done right — 49 structured workflows, persistent memory across sessions, integrated learning partner, and impeccable UI design system. Works with Claude Code, Windsurf, Cursor, Gemini CLI, OpenCode, and Codex.", "name": "Favio Vazquez",
