Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add joyehuang/minimind-notes --skill minimind-learning-skillgit clone --depth 1 https://github.com/joyehuang/minimind-notesWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/joyehuang/minimind-notes/minimind-learning-skill)<a href="https://agentmods.dev/skills/joyehuang/minimind-notes/minimind-learning-skill"><img src="https://agentmods.dev/badge/skills/joyehuang/minimind-notes/minimind-learning-skill.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00076 | $0.04343 |
| Opus 5 | $0.00038 | $0.02171 |
| Sonnet 5 | $0.00015 | $0.00869 |
| Haiku 4.5 | $0.00008 | $0.00434 |
Grade A, and why
minimind-learning scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 587 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MiniMind Learning Assistant
自动化学习笔记系统,静默记录你的 MiniMind 学习历程。
When to use
自动激活场景:
-
学习开始时:
- 用户说:"开始今天的学习"、"开始学习"、"今天学什么"
- 用户说:"继续学习"、"学习 MiniMind"
-
讨论 MiniMind 内容时:
- 提问包含:RMSNorm, LayerNorm, RoPE, YaRN, Attention, GQA, SwiGLU, Transformer, LoRA, DPO, PPO, GRPO, SFT, RLHF, RLAIF, MoE
- 问题词:什么是、如何、为什么、怎样、解释、原理
- 遇到问题:报错、错误、失败、Bug
-
显式记录请求:
- 用户说:"记录一下"、"保存笔记"、"写入笔记"
How to use
1. 初始化检查
首次激活时,确保笔记结构存在:
# 检测 Git 仓库根目录
git rev-parse --show-toplevel
# 验证是 MiniMind 仓库(存在以下文件)
# - model/model_minimind.py
# - trainer/train_pretrain.py
# - README.md (包含 "MiniMind")
# 创建笔记目录(如不存在)
mkdir -p docs/learning_materials
# 从模板初始化文件(如不存在)
# - docs/notes.md
# - docs/learning_log.md
# - docs/knowledge_base.md
# - docs/learning_materials/README.md
模板位置:~/.claude/skills/minimind-learning/templates/
2. 学习引导模式
当用户说"开始学习"时,主动引导:
👋 欢迎开始今天的 MiniMind 学习!
你想学习哪个模块?
**基础组件**:
1. 归一化技术 - RMSNorm, LayerNorm
2. 位置编码 - RoPE, YaRN
3. 注意力机制 - Attention, GQA
4. 前馈网络 - FeedForward, SwiGLU
**训练技术**:
5. 预训练 - Pretraining
6. 监督微调 - SFT
7. 参数高效微调 - LoRA
8. 强化学习 - DPO, PPO, GRPO
直接告诉我编号或主题名称,我会为你讲解!
(学习过程中我会自动记录笔记到 `docs/` 目录)
3. 触发检测
每次用户消息后,检查是否满足以下任一条件:
Tier 1: 即时触发(立即更新笔记)
# MiniMind 术语(50+)
TERMS = [
# 架构
"RMSNorm", "LayerNorm", "BatchNorm", "GroupNorm",
"RoPE", "YaRN", "ALiBi", "位置编码",
"Attention", "注意力", "GQA", "MQA", "FlashAttention",
"FeedForward", "前馈", "SwiGLU", "GELU", "GLU",
"Transformer", "TransformerBlock",
# 训练
"pretrain", "预训练", "pretraining",
"SFT", "监督微调", "fine-tuning", "微调",
"LoRA", "lora", "LoRA-r", "LoRA-alpha",
"DPO", "PPO", "GRPO", "SPO",
"RLHF", "RLAIF", "强化学习",
"distillation", "蒸馏", "知识蒸馏",
# 模型
"MiniMind", "MiniMind-Dense", "MiniMind-MoE",
"MoE", "混合专家", "expert routing",
"MiniMind-Reason", "R1",
]
# 问题词
QUESTION_WORDS = ["什么是", "如何", "为什么", "怎样", "解释", "原理", "作用"]
# 问题指示
PROBLEM_MARKERS = ["报错", "错误", "失败", "Bug", "不工作", "问题"]
# 检查
if any(term in user_message for term in TERMS):
trigger_tier_1()
elif any(word in user_message for word in QUESTION_WORDS):
trigger_tier_1()
elif any(marker in user_message for marker in PROBLEM_MARKERS):
trigger_tier_1()
What ships with it
10 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- .gitignore 243 B
- .minimind-learning.json.example 356 B
- LICENSE 1.0 KB
- scripts/init-for-learning.bat 3.1 KB runs code
- scripts/init-for-learning.sh 3.0 KB runs code
- scripts/validate_notes.py 8.9 KB runs code
- templates/knowledge_base.md.template 7.5 KB
- templates/learning_log.md.template 1.1 KB
- templates/learning_materials_readme.md.template 5.1 KB
- templates/notes.md.template 4.3 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 587 lines · 76 tokens per session scan A 8acecdd0bffb
minimind-learning is a skill published in the GitHub repository joyehuang/minimind-notes (184 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 76 tokens to every session and 4,343 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
ai-engineering-curriculum
Use when structured AI engineering curriculum — 382 skills + 99 prompts across 20 phases covering ML, deep learning, LLMs, agents, and production systems. Use when learning AI, building AI skills,.
andrej-karpathy
Agente que simula Andrej Karpathy — ex-Director of AI da Tesla, co-fundador da OpenAI, fundador da Eureka Labs, e o maior educador de deep learning do mundo.
andrej-karpathy
Agente que simula Andrej Karpathy — ex-Director of AI da Tesla, co-fundador da OpenAI, fundador da Eureka Labs, e o maior educador de deep learning do mundo.
andrej-karpathy
Agente que simula Andrej Karpathy — ex-Director of AI da Tesla, co-fundador da OpenAI, fundador da Eureka Labs, e o maior educador de deep learning do mundo.
outlines
Outlines: structured JSON/regex/Pydantic LLM generation.
guidance
Constrain LLM output with grammars; guarantee valid JSON.