claude-english-immersion

KO | EN

gitlite — search

#python #react #docker #nodejs #python3 #web #django #java #javascript #hacktoberfest #django-rest-framework #express

claude-english-immersion

★ 69 Open GitHub ↗

Claude Code skills for passive English immersion + PTE exam prep. Turn daily AI conversations into language learning.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Claude English Immersion + PTE Skills

Turn your daily Claude Code sessions into passive English immersion and PTE (Pearson Test of English) exam prep.

> Created by [Roland Wayne](https://github.com/rolandwonglonam). The skill architecture, learning methodology, error taxonomy, PTE absorption mechanism, and Wittgenstein-style teaching framework were designed and iterated by Roland through daily use. Claude assisted with implementation and documentation.

**[中文说明](#中文说明) | [English](#what-this-is)**

---

## What This Is

Two Claude Code skills that work together:

- **english-immersion** — The runtime. When English Mode is ON, Claude replies in English, silently logs your errors, teaches unknown words Wittgenstein-style (usage, not translation), and delivers batch error reviews at session end. No per-sentence nagging.
- **pte** — The feeder. Ingest PTE teacher transcripts (YouTube, etc.), extract structured rules, and write them to a methodology file. english-immersion reads these rules passively and demonstrates them in its replies — you absorb PTE techniques without active study.

## Core Idea

If you already talk to Claude every day, switching that conversation to English gives you hours of passive input per day without adding anything to your schedule. This skill makes that switch frictionless.

## Features

- Binary ON/OFF state machine (say "english mode on" / "english mode off")
- Input logging (append-only, English/mixed inputs only)
- Error tagging with pattern taxonomy (e.g., `missing-be-verb`, `preposition-about-vs-with`)
- Wittgenstein-style word teaching: usage + example + contrast, no translation
- Session-end batch error review (not inline correction)
- Weekly reports with error frequency, new words, PTE progress
- PTE Absorption Hook: passive rule demonstration in Claude's replies
- Active PTE practice sub-commands: `/pte-essay`, `/pte-summarize`, `/pte-describe`, `/pte-read-aloud`
- Sliding anchor: exam date proximity drives demonstration density

## Installation

### 1. Copy skills to Claude Code

```bash
# Copy both skill directories to your Claude Code skills folder
cp -r english-immersion ~/.claude/skills/
cp -r pte ~/.claude/skills/
```

### 2. Set up data directory

Copy the data template to wherever you keep persistent files (e.g., your Obsidian vault, a notes folder, etc.):

```bash
cp -r data-template ~/path/to/your/data/english-immersion/
```

### 3. Update paths in SKILL.md

Edit both `SKILL.md` files and replace `data/` with the actual path to your data directory:

```
# In english-immersion/SKILL.md and pte/SKILL.md
# Replace all occurrences of:
data/
# With your actual path, e.g.:
~/notes/english-immersion/
```

### 4. Customize context

Edit `english-immersion/SKILL.md` line 10 — replace the placeholder with your actual context:

```
**Context**: [Your location]. Target: PTE [score]+ in all four sections. Baseline: [your baseline]. Timeline: [your timeline].
```

## Usage

### English Mode

```
"english mode on"    → Claude switches to English replies + starts logging
"english mode off"   → Back to normal
```

### PTE Feeding

```
/pte feed <paste a PTE teacher transcript>   → Extract rules to methodology.md
/pte rule <one rule in plain text>           → Manually add a rule
/pte anchor                                  → Set exam date + current/target score
/pte vocab <collocation list>                → Add high-frequency collocations
/pte status                                  → View progress summary
```

### PTE Practice (optional, pull-only)

```
/pte-essay [topic]       → 200-300 word argumentative essay
/pte-summarize [text]    → Single-sentence summary (SWT task)
/pte-describe [image]    → 40-second image description (DI task)
/pte-read-aloud [text]   → Calibrated read-aloud passage (RA task)
```

## Architecture

```
english-immersion/          ← Runtime skill (always-on when mode: on)
├── SKILL.md                   Main spec + PTE Absorption Hook
├── reference/
│   ├── wittgenstein-style.md  Word teaching methodology
│   ├── error-pattern-taxonomy.md  Error classification system
│   └── pte-rubrics.md        PTE scoring guides
└── sub-skills/
    ├── pte-essay.md           Write Essay practice
    ├── pte-summarize.md       Summarize Written Text practice
    ├── pte-describe.md        Describe Image practice
    └── pte-read-aloud.md      Read Aloud practice

pte/                        ← Feeder skill (fire-and-forget)
├── SKILL.md                   Ingest transcripts, extract rules
└── reference/
    ├── structure-templates.md Rule templates per PTE task
    ├── exam-checklist.md      Exam day checklist
    └── design-dbs-notes.md    Design decisions archive

data/                       ← Your persistent data (lives in your vault/notes)
├── state.md                   ON/OFF + sliding anchor
├── inputs_log.md              Raw English inputs (append-only)
├── errors_tagged.md           Tagged errors with patterns
├── words_learned.md           Wittgenstein-style word cards
├── methodology.md             PTE rules (written by /pte, read by english-immersion)
├── absorption_log.md          Rule hit stream (written by english-immersion)
├── pte_vocab.md               Academic Collocation List
├── pte_progress.md            Practice scores over time
└── weekly_reports/            Weekly summaries
```

## How the Two Skills Work Together

1. **pte** writes rules to `methodology.md` (from YouTube transcripts or manual input)
2. **english-immersion** reads `methodology.md` during ON mode
3. When you type English, english-immersion checks if your input matches any learning rules → logs hits to `absorption_log.md`
4. When Claude replies, it silently demonstrates the least-hit rules in natural phrasing
5. After 3 spontaneous hits, a rule graduates to "mastered"
6. Weekly reports track absorption progress

---

## 中文说明

### 这是什么

两个 Claude Code skill，把你每天跟 Claude 对话的时间变成英语沉浸式学习 + PTE 备考。

核心逻辑很简单：你本来就每天跟 Claude 聊天，把对话语言切成英文，每天就多了几个小时的被动输入，不需要额外花时间。

### 为什么做这个

Vibe coding 时代，开发者每天跟 Claude Code 对话好几个小时。这些时间本来就在花，何不顺便把英语练了？不需要额外下 App、不需要专门腾时间，把对话语言切成英文就行。

关键设计决策：
- **不逐句纠错**。每句都纠正会让人烦到关掉。错误静默记录，session 结束时批量复盘
- **维特根斯坦式教词**。不给中文翻译，用「什么场景下用」+「一个例句」+「跟近义词的区别」来教。记得更牢
- **PTE 规则被动吸收**。从 YouTube 老师视频里提取考试技巧，Claude 在日常回复里悄悄示范这些技巧，你不知不觉就学会了
- **考试日期越近，示范越密**。滑动锚点机制，离考试越近 Claude 回复里塞的 PTE 技巧越多

### 安装

1. 把 `english-immersion/` 和 `pte/` 复制到 `~/.claude/skills/`
2. 把 `data-template/` 复制到你的笔记目录（Obsidian vault 或任何持久化位置）
3. 修改两个 `SKILL.md` 里的 `data/` 路径，指向你实际的数据目录
4. 在 `english-immersion/SKILL.md` 第 10 行填入你的个人信息（所在地、目标分、当前水平、时间线）

### 使用

```
"english mode on"     → 开启英语模式
"english mode off"    → 关闭
/pte feed <文稿>      → 喂 PTE 老师的视频文稿，自动提取规则
/pte status           → 查看吸收进度
/pte-essay [话题]     → 主动练习写作（可选，不强制）
```

### 作者

由 [Roland Wayne](https://github.com/rolandwonglonam) 设计和创造。整个 skill 的架构设计、学习方法论、错误分类体系、PTE 被动吸收机制、维特根斯坦式教学框架，都是 Roland 在日常使用中反复迭代出来的。Claude 负责实现和文档化。

---

## Roadmap / 开发预留

This is a functional but intentionally incomplete release. The architecture is designed to be extended. Here are some directions you can take it:

这是一个可用但刻意未做完的版本，架构预留了扩展空间。你可以根据自己的需要继续开发：

- **More language support** — The framework is language-agnostic. Swap PTE for IELTS, TOEFL, JLPT, DELF, or any other exam by replacing the rubrics and methodology rules
- **Listening / speaking integration** — Current limitation: Claude can't hear audio. If you find a way to pipe speech-to-text into the conversation, the error taxonomy and absorption mechanism work out of the box
- **Spaced repetition** — `words_learned.md` is a flat file. Wire it into Anki or any SRS system for better retention
- **Multi-user** — Currently single-user. Add a user ID to the data files to support teams or classrooms
- **Analytics dashboard** — The data is all in Markdown. Build a visualization layer on top of `errors_tagged.md` and `absorption_log.md`
- **Auto-generated practice** — Use the error patterns to automatically generate targeted exercises (e.g., if `missing-be-verb` keeps recurring, generate fill-in-the-blank drills)

PRs and forks welcome.

---

## License

MIT

claude-english-immersion

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README