Back to Details

teach-oh

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

/teach-oh

/teach-oh

Setup that gathers project context and saves it to AGENTS.md (the cross-agent standard). Run when starting on a project or when strategy/aims have shifted.
用于收集项目上下文并将其保存到AGENTS.md（跨Agent标准文件）的初始化流程。在开始新项目或项目策略/目标发生变化时运行。

When to Use

使用时机

Invoke 
/teach-oh
when:
  • Starting on a new project - Before diving into work
  • Context keeps getting lost - AI assistants don't "get" your project
  • Onboarding a new AI tool - Establish shared understanding upfront
  • After major strategic shifts - When aims or constraints have changed
Do not use when: You're mid-task. This is setup, not execution.
在以下场景中调用
/teach-oh
：
  • 启动新项目时 - 在投入工作之前
  • 上下文频繁丢失时 - AI助手无法理解你的项目
  • 接入新AI工具时 - 提前建立统一认知
  • 重大战略调整后 - 当项目目标或约束条件发生变化时
请勿在以下场景使用：任务进行中。这是初始化流程，而非执行环节。

The Process

操作流程

Step 1: Explore the Codebase

步骤1：探索代码库

Before asking questions, scan the project independently:
Structure & Stack
  • Directory layout, key folders
  • Package files (package.json, Cargo.toml, go.mod, etc.)
  • Build configuration, CI/CD setup
Existing Context
  • AGENTS.md, CLAUDE.md, README, CONTRIBUTING
  • .oh/
    , 
    docs/adr/
    directories
  • Any existing project documentation
Patterns & Conventions
  • Naming conventions in code
  • File organization patterns
  • Recent git commits for style and focus
Don't ask about what you can discover. Use exploration to form better questions.
在提问前，先自主扫描项目：
结构与技术栈
  • 目录布局、关键文件夹
  • 包管理文件（package.json、Cargo.toml、go.mod等）
  • 构建配置、CI/CD设置
现有上下文
  • AGENTS.md、CLAUDE.md、README、CONTRIBUTING文档
  • .oh/
    、
    docs/adr/
    目录
  • 任何已有的项目文档
模式与规范
  • 代码命名规范
  • 文件组织模式
  • 近期git提交记录，了解代码风格与当前工作重点
不要询问可自行发现的内容。通过探索来提出更有质量的问题。

Step 2: Ask About What Matters

步骤2：询问关键信息

After exploration, ask targeted questions about what couldn't be inferred. Focus on strategy and aims, not just code.
探索完成后，针对无法自行推断的内容提出针对性问题。重点关注策略与目标，而非仅局限于代码细节。

Purpose & Aims

项目目标与定位

  • What is this project trying to achieve?
  • Who uses it? What change in their behavior indicates success?
  • What's the current focus or priority?
  • 该项目的核心目标是什么？
  • 目标用户是谁？哪些行为变化代表项目成功？
  • 当前的工作重点或优先级是什么？

Strategic Constraints

战略约束条件

  • What constraints never bend? (Compliance, performance SLAs, etc.)
  • What trade-offs has the team made intentionally?
  • What's explicitly out of scope?
  • 哪些约束是绝对不可违背的？（合规要求、性能SLA等）
  • 团队曾主动做出过哪些权衡取舍？
  • 哪些内容明确不属于项目范围？

Team & Decision-Making

团队与决策机制

  • How does the team make decisions?
  • What does "done" mean here?
  • Any patterns or practices that are sacred?
  • 团队的决策流程是怎样的？
  • 这里的“完成”标准是什么？
  • 有没有哪些模式或实践是团队坚守的？

What to Avoid

需要规避的内容

  • Past mistakes that shouldn't be repeated
  • Patterns that look tempting but don't fit
  • Areas of the codebase that are sensitive
Ask only what you couldn't discover. Respect the user's time.
  • 不应重复的过往错误
  • 看似诱人但不适合项目的模式
  • 代码库中哪些部分是敏感区域？
只询问无法自行发现的内容，尊重用户的时间。

Step 3: Write Persistent Context

步骤3：写入持久化上下文

Synthesize findings into a structured section and offer to append to AGENTS.md (or create it if missing).
Offer two sections:
  1. Open Horizons Framework - The strategic framework for AI-assisted work (offer to include if user uses OH skills)
  2. Project Context - Project-specific aims, constraints, and patterns (always include)
markdown
undefined
将调研结果整理为结构化内容，并询问是否追加到AGENTS.md中（若文件不存在则创建）。
提供两个核心板块：
  1. Open Horizons Framework - AI辅助工作的战略框架（若用户使用OH技能则提供）
  2. 项目上下文 - 项目专属的目标、约束与模式（必选）
markdown
undefined

Open Horizons Framework

Open Horizons Framework

The shift: Action is cheap. Knowing what to do is scarce.
The sequence: aim → problem-space → problem-statement → solution-space → execute → ship
Where to start (triggers):
  • Can't explain why you're building this → 
    /aim
  • Keep hitting the same blockers → 
    /problem-space
  • Solutions feel forced → 
    /problem-statement
  • About to start coding → 
    /solution-space
  • Work is drifting or reversing → 
    /salvage
Reflection skills (use anytime):
  • /review
    - Check alignment before committing
  • /dissent
    - Seek contrary evidence before one-way doors
  • /salvage
    - Extract learning, restart clean
Key insight: Enter at the altitude you need. Climb back up when you drift.
核心转变： 执行成本低廉，但明确方向才是稀缺资源。
工作流程： 目标定义 → 问题域分析 → 问题陈述 → 解决方案域 → 执行 → 交付
启动触发点：
  • 无法解释构建该项目的原因 → 
    /aim
  • 反复遇到相同障碍 → 
    /problem-space
  • 解决方案显得生硬牵强 → 
    /problem-statement
  • 即将开始编码 → 
    /solution-space
  • 工作方向偏离或出现倒退 → 
    /salvage
反思类技能（随时可用）：
  • /review
    - 提交前检查是否对齐目标
  • /dissent
    - 做出不可逆决策前寻求相反证据
  • /salvage
    - 提取经验教训，重新开始
核心洞察： 从合适的阶段介入，偏离时及时回归顶层目标。

Project Context

项目上下文

Purpose

项目定位

[What this project does and why it matters]
[该项目的功能与价值]

Current Aims

当前目标

[Active goals and what success looks like]
[活跃目标及成功衡量标准]

Key Constraints

关键约束

  • [Constraint 1]: [Why it exists]
  • [Constraint 2]: [Why it exists]
  • [约束1]：[存在原因]
  • [约束2]：[存在原因]

Patterns to Follow

需遵循的模式

  • [Pattern]: [When/why to use it]
  • [模式]：[适用场景/原因]

Anti-Patterns to Avoid

需规避的反模式

  • [Anti-pattern]: [Why it's problematic here]
  • [反模式]：[在本项目中的问题]

Decision Context

决策上下文

[How the team makes decisions, what "done" means]
undefined
[团队决策方式、“完成”的定义]
undefined

Output Format

输出格式

Present findings for user approval before writing:
undefined
在写入前先向用户展示调研结果以获得确认：
undefined

Teach-OH Summary

Teach-OH 总结

What I Discovered

调研发现

  • Stack: [technologies found]
  • Structure: [key patterns observed]
  • Existing context: [what docs already exist]
  • 技术栈：[已发现的技术]
  • 项目结构：[观察到的关键模式]
  • 现有上下文：[已存在的文档]

Questions I Have

待确认问题

  1. [Question about aims/strategy]
  2. [Question about constraints]
  3. [Question about team practices]
After your answers, I'll draft a AGENTS.md section for your approval.

After user answers:
  1. [关于目标/策略的问题]
  2. [关于约束条件的问题]
  3. [关于团队实践的问题]
在您解答后，我将起草AGENTS.md的内容供您确认。

用户解答后：

Proposed AGENTS.md Addition

拟写入AGENTS.md的内容

[Draft the Project Context section]
Write to AGENTS.md? [Yes/No - show what will be added/updated]
undefined
[起草项目上下文板块]
是否写入AGENTS.md？ [说明将添加/更新的内容]
undefined

Example

示例

/teach-oh
Agent explores, then:
undefined
/teach-oh
Agent完成探索后：
undefined

Teach-OH Summary

Teach-OH 总结

What I Discovered

调研发现

  • Stack: TypeScript, React, Node.js, PostgreSQL
  • Structure: Monorepo with apps/ and packages/
  • Existing context: README exists, no AGENTS.md or CLAUDE.md
  • Recent focus: Auth refactor (last 12 commits)
  • 技术栈：TypeScript、React、Node.js、PostgreSQL
  • 项目结构：单仓库架构，包含apps/与packages/目录
  • 现有上下文：存在README文档，无AGENTS.md或CLAUDE.md
  • 近期工作重点：认证流程重构（最近12次提交）

Questions I Have

待确认问题

  1. What's the main aim right now? The auth work—is that about security, UX, or maintenance?
  2. I see both REST and GraphQL endpoints. Is one preferred, or context-dependent?
  3. Any performance constraints? I see a Redis layer but no docs on when to use it.
  4. What patterns should I follow vs. avoid?

**User answers, agent drafts:**
  1. 当前的核心目标是什么？认证工作是为了提升安全性、用户体验还是可维护性？
  2. 我看到项目同时存在REST与GraphQL接口，是否有偏好场景，还是根据上下文选择？
  3. 有没有性能约束？我看到Redis层但没有相关使用文档。
  4. 我应该遵循哪些模式，规避哪些模式？

**用户解答后，Agent起草内容：**

Proposed AGENTS.md Addition

拟写入AGENTS.md的内容

Open Horizons Framework

Open Horizons Framework

The shift: Action is cheap. Knowing what to do is scarce.
The sequence: aim → problem-space → problem-statement → solution-space → execute → ship
Where to start (triggers):
  • Can't explain why you're building this → 
    /aim
  • Keep hitting the same blockers → 
    /problem-space
  • Solutions feel forced → 
    /problem-statement
  • About to start coding → 
    /solution-space
  • Work is drifting or reversing → 
    /salvage
Reflection skills: 
/review
, 
/dissent
, 
/salvage
Key insight: Enter at the altitude you need. Climb back up when you drift.
核心转变： 执行成本低廉，但明确方向才是稀缺资源。
工作流程： 目标定义 → 问题域分析 → 问题陈述 → 解决方案域 → 执行 → 交付
启动触发点：
  • 无法解释构建该项目的原因 → 
    /aim
  • 反复遇到相同障碍 → 
    /problem-space
  • 解决方案显得生硬牵强 → 
    /problem-statement
  • 即将开始编码 → 
    /solution-space
  • 工作方向偏离或出现倒退 → 
    /salvage
反思类技能： 
/review
、
/dissent
、
/salvage
核心洞察： 从合适的阶段介入，偏离时及时回归顶层目标。

Project Context

项目上下文

Purpose

项目定位

Customer portal for B2B SaaS. Users are account admins managing their organization's settings, users, and billing.
面向B2B SaaS的客户门户，用户为企业管理员，用于管理组织设置、用户与账单。

Current Aims

当前目标

  • Simplify auth flow so new SSO providers can be added in <1 day (currently takes 2 weeks)
  • Success: Adding Okta support by end of Q1
  • 简化认证流程，使新增SSO提供商的时间从2周缩短至1天以内
  • 成功标志：在Q1末完成Okta支持的集成

Key Constraints

关键约束

  • SOC2 compliance: All auth changes need security review
  • <200ms API response time (SLA commitment)
  • No breaking changes to existing OAuth integrations
  • SOC2合规要求：所有认证变更需经过安全审核
  • API响应时间<200ms（SLA承诺）
  • 不得对现有OAuth集成造成破坏性变更

Patterns to Follow

需遵循的模式

  • REST for public API, GraphQL for internal dashboard
  • Redis for session state only, not general caching
  • Feature flags via LaunchDarkly for gradual rollouts
  • 公共API使用REST，内部仪表盘使用GraphQL
  • Redis仅用于会话状态存储，不用于通用缓存
  • 使用LaunchDarkly的功能标志进行逐步发布

Anti-Patterns to Avoid

需规避的反模式

  • Don't abstract OAuth providers into single interface (they have semantic differences)
  • Avoid adding new environment variables without updating .env.example
  • 不要将OAuth提供商抽象为单一接口（它们存在语义差异）
  • 添加新环境变量时，必须同步更新.env.example

Decision Context

决策上下文

Tech lead approves architecture changes. PRs need one review. "Done" = deployed to staging with passing E2E tests.
Write to AGENTS.md? This will create a new file at ./AGENTS.md
undefined
架构变更需经技术负责人批准。PR需要至少一次审核。“完成”的定义：部署到预发布环境且E2E测试全部通过。
是否写入AGENTS.md？ 将在./AGENTS.md路径下创建新文件
undefined

What This Enables

带来的价值

With project context established:
  • /aim
    frames outcomes in your language
  • /problem-space
    knows which constraints are real
  • /dissent
    understands your risk tolerance
  • /review
    checks against your definition of done
  • /execute
    follows your patterns
建立项目上下文后：
  • /aim
    将用您的语言定义成果
  • /problem-space
    了解哪些约束是真实存在的
  • /dissent
    理解您的风险承受能力
  • /review
    对照您的“完成”标准进行检查
  • /execute
    遵循您的项目模式

Notes

注意事项

  • Context lives in AGENTS.md so it persists across sessions
  • Keep it focused—this isn't documentation, it's working context
  • Re-run when aims shift, constraints change, or context feels stale
Remember: This is setup, not ongoing work. Invest 10 minutes when starting or when things have shifted.
  • 上下文存储在AGENTS.md中，可跨会话持久化
  • 保持内容聚焦——这不是正式文档，而是工作上下文
  • 当目标变更、约束调整或上下文过时，重新运行该流程
请记住： 这是初始化流程，而非日常工作。在项目启动或发生重大变化时投入10分钟完成设置。