Free to read · no account needed
Learning paths
Learn by scenario
For absolute beginners. Small steps, jargon explained, open to read. ~119 guides live—including new starter packs, paths, and trust habits.
面向完全新手:小步骤、术语会解释、全文可直接读。已上线 119 篇——含新提示词包、路径与核实习惯。
Start here
- Accounts, safety, first prompts
- Your first 7 days (~14 min/day)
- Copy-paste prompt packs
- 10 beginner mistakes + fixes
- When AI sounds sure but wrong
- What AI chat can and cannot do
- Goal · Format · Boundaries
- How to choose a model (2026)
Paths (by job)
- Writing & expression
- Research & learning
- Code & productivity
- Career (resume · letter · interview)
- Study (explain · quiz · plan)
- Office weekly routine
System track (skills · agents · plugins)
- What is a skill?
- Write your first skill
- Skills vs GPTs vs Projects
- Chat vs assistants vs agents
- Define done + permissions
- Plan → act → check
- Skills + MCP + plugins stack
- MCP in plain words
- Connect a safe first server
- Human-in-the-loop checkpoints
- Research agent pattern
- Writing pipeline agent
- Coding agent boundaries
- Failure modes and retries
- Logging what matters
- Subagents and handoffs
- Team playbook
- From one-shot to scheduled
- Style guides that stick
- PR descriptions that help
- Secure coding prompts
- Drive / mail / calendar connectors
- Multimodal basics
- Debug when it ignores you
- Version your prompt templates
- Stakeholder update
- Fact-check social claims
- Debug broken MCP calls
- Install and evaluate plugins
- Agent blast radius
New: craft, safety, product
- Personal AI stack in one week
- Eval your own outputs
- MCP tools permissions
- Plugin directory hygiene
- Package plugins for teammates
- Build vs buy plugins
- Multilingual without muddy mix
- Role without cosplay
- Rubrics and acceptance tests
- Competitor one-pager
- Interview prep brief
- Literature map fast
- Copyright and your voice
- Skill anti-patterns
- Explicit vs auto invoke
- Pack references and scripts
- Port skills across hosts
- Team shared skill library
- Test a skill before trust
- Skill versioning
- Browser vs API vs MCP
- Landing page copy loop
- Newsletter in 45 min
- Translate then localize
- Cover letter + mock interview
- Tests with the model
- Negative constraints that work
- Claude Cowork multi-step
- Computer use safely
- Reject AI slop
- Disclose AI assistance
- Least-privilege checklist
- Ship a static guide site
- HK bilingual workday
- Voice & mobile workflows
- Resume without inventing
- Study loop: explain → quiz → plan
More foundations & craft
- Context, memory, Projects
- Standing instructions vs each message
- When to switch models
- Few-shot examples that help
- Tables & simple JSON
- Draft → critique → revise
- Source tiers
- Claim–evidence matrix
- Outline before prose
- Voice consistency pass
- Minimal repro first
- Explain then change (before the patch)
- What never to paste
- Hide personal data (PII)
- Free vs paid plans
- Accounts and safety first
Tool-track picks
- Clear everyday ChatGPT prompts
- Draft & revise with ChatGPT
- Research with a plan
- Meeting notes → action
- Email: correct → polished
- Verify before you share
- Live topic digest
- How to choose a model
- ChatGPT Projects as workspaces
- Claude Projects knowledge
- Trend to one-page brief
- Deep research mode
- Canvas & iterative docs
- Custom GPTs for repeat work
- Data analysis with files
- Image prompts that ship
- Artifacts for deliverables
- Code review checklist
- When to use extended thinking
- Long document pass
- Informal brainstorm
- Tone for social drafts
- Grok vs search engines
- X context without echo chamber
Full ChatGPT / Claude / Grok tracks: Tools.
All 119 guides are listed above, grouped by scenario. To browse by tool instead, see the Tools page.
从这里开始
- 账号、安全与第一批提示词
- 头 7 天上手(每天约 14 分钟)
- 可复制提示词包
- 新手十大坑与改法
- 听起来很有把握却是错的
- AI 聊天能做什么、不能做什么
- 目标 · 格式 · 边界
- 怎么选模型(2026)
按任务走的路径
系统向(技能 · 智能体 · 插件)
- 什么是技能?
- 写你的第一个技能
- Skill / GPT / Project 怎么选
- 聊天、助手与智能体
- 先定义「做完」和权限
- 计划 → 行动 → 检查
- Skill + MCP + 插件这一叠
- MCP 白话解释
- 安全地连第一个
- 人工把关检查点
- 调研型智能体套路
- 写作流水线智能体
- 编程智能体边界
- 失败模式与重试
- 记该记的运行日志
- 子智能体与交接
- 团队玩法手册
- 从一次跑通到定时
- 能站住的风格指南
- 帮得上忙的 PR 描述
- 安全向编程提示
- 网盘 / 邮件 / 日历连接器
- 多模态基础
- 模型不理你时怎么调
- 可版本化的提示词模板
- 干系人进展更新
- 社交声称核实
- 排查坏掉的 MCP 调用
- 安装并评估插件
- 智能体影响范围
新增:写法、安全、产品
- 一周搭好个人 AI 组合
- 一分钟验自己的输出
- MCP 工具权限与批准
- 插件目录卫生
- 给队友打包插件
- 插件自建还是采购
- 多语言别搅成沙拉
- 角色设定别玩 cosplay
- 量表与验收测试
- 竞品一页纸
- 访谈准备简报
- 文献地图速绘
- 版权与训练你的语气
- 技能反模式
- 显式调用 vs 自动
- 打包引用与脚本
- 跨产品移植技能
- 团队共享技能库
- 信任前先测技能
- 技能版本与变更日志
- 浏览器 vs API vs MCP
- 落地页文案循环
- 45 分钟写完通讯
- 先翻译再本地化
- 求职信 + 模拟面试
- 用模型写小测试
- 管用的禁止项
- Claude Cowork 多步任务
- 安全使用电脑控制
- 拒绝 AI 注水腔
- 披露 AI 协助
- 最小权限清单
- 上线静态指南站
- 港式双语工作日
- 语音与手机工作流
- 不编造的简历改写
- 学习循环:讲解→自测→计划
更多基础与写法
- 上下文、记忆、Projects
- 固定说明 vs 每句话
- 什么时候换模型
- 少样本例子怎么给
- 表格与简单 JSON
- 草稿 → 挑剔 → 修改
- 来源分级
- 主张–证据表
- 先提纲后正文
- 语气一致检查
- 先做最小复现
- 先解释再改代码
- 哪些东西不要粘贴
- 隐藏个人数据
- 免费 vs 付费方案
- 账号与安全先行
分工具精选
- 日常清晰 ChatGPT 提示
- 用 ChatGPT 起草与改稿
- 有计划的研究
- 会议笔记 → 行动
- 邮件:从正确到得体
- 分享前核验
- 实时话题摘要
- 如何选模型
- ChatGPT Projects 当工作区
- Claude Projects 知识库
- 热点收成一页简报
- 深度研究模式
- Canvas 与迭代文档
- 自定义 GPT 做重复活
- 带文件做数据分析
- 能交差的图像提示
- 交付物 Artifacts
- 代码审查清单
- 何时用扩展思考
- 长文档一遍过
- 非正式头脑风暴
- 社交草稿语气
- Grok 与搜索引擎
- X 语境避免回音室
完整 ChatGPT / Claude / Grok 分工具列表见 Tools。
上面按场景列出了全部 119 篇教程;想按工具来找,可以看 Tools 页。