mirror of
https://github.com/sweetwisdom/everything-claude-code-zh.git
synced 2026-03-22 06:20:10 +00:00
docs: 完成所有文档的中文翻译并应用到项目
This commit is contained in:
xuxiang
@@ -1,23 +1,23 @@
|
||||
---
|
||||
name: continuous-learning
|
||||
description: Automatically extract reusable patterns from Claude Code sessions and save them as learned skills for future use.
|
||||
description: 自动从 Claude Code 会话（Sessions）中提取可重用的模式，并将其保存为学习到的技能以供未来使用。
|
||||
---
|
||||
|
||||
# Continuous Learning Skill
|
||||
# 持续学习技能（Continuous Learning Skill）
|
||||
|
||||
Automatically evaluates Claude Code sessions on end to extract reusable patterns that can be saved as learned skills.
|
||||
在会话结束时自动评估 Claude Code 会话（Sessions），以提取可保存为学习技能（Learned Skills）的可重用模式。
|
||||
|
||||
## How It Works
|
||||
## 工作原理
|
||||
|
||||
This skill runs as a **Stop hook** at the end of each session:
|
||||
该技能作为 **停止钩子（Stop hook）** 在每个会话结束时运行：
|
||||
|
||||
1. **Session Evaluation**: Checks if session has enough messages (default: 10+)
|
||||
2. **Pattern Detection**: Identifies extractable patterns from the session
|
||||
3. **Skill Extraction**: Saves useful patterns to `~/.claude/skills/learned/`
|
||||
1. **会话评估（Session Evaluation）**：检查会话是否有足够的消息（默认：10 条以上）
|
||||
2. **模式检测（Pattern Detection）**：识别会话中可提取的模式
|
||||
3. **技能提取（Skill Extraction）**：将有用的模式保存到 `~/.claude/skills/learned/`
|
||||
|
||||
## Configuration
|
||||
## 配置
|
||||
|
||||
Edit `config.json` to customize:
|
||||
编辑 `config.json` 进行自定义：
|
||||
|
||||
```json
|
||||
{
|
||||
@@ -40,19 +40,19 @@ Edit `config.json` to customize:
|
||||
}
|
||||
```
|
||||
|
||||
## Pattern Types
|
||||
## 模式类型
|
||||
|
||||
| Pattern | Description |
|
||||
| 模式（Pattern） | 描述（Description） |
|
||||
|---------|-------------|
|
||||
| `error_resolution` | How specific errors were resolved |
|
||||
| `user_corrections` | Patterns from user corrections |
|
||||
| `workarounds` | Solutions to framework/library quirks |
|
||||
| `debugging_techniques` | Effective debugging approaches |
|
||||
| `project_specific` | Project-specific conventions |
|
||||
| `error_resolution` | 特定错误的解决方式 |
|
||||
| `user_corrections` | 来自用户修正的模式 |
|
||||
| `workarounds` | 框架/库特有问题的变通方案 |
|
||||
| `debugging_techniques` | 有效的调试方法 |
|
||||
| `project_specific` | 项目特定的约定 |
|
||||
|
||||
## Hook Setup
|
||||
## 钩子设置（Hook Setup）
|
||||
|
||||
Add to your `~/.claude/settings.json`:
|
||||
添加到你的 `~/.claude/settings.json`：
|
||||
|
||||
```json
|
||||
{
|
||||
@@ -68,43 +68,43 @@ Add to your `~/.claude/settings.json`:
|
||||
}
|
||||
```
|
||||
|
||||
## Why Stop Hook?
|
||||
## 为什么使用停止钩子（Stop Hook）？
|
||||
|
||||
- **Lightweight**: Runs once at session end
|
||||
- **Non-blocking**: Doesn't add latency to every message
|
||||
- **Complete context**: Has access to full session transcript
|
||||
- **轻量级（Lightweight）**：在会话结束时运行一次
|
||||
- **非阻塞（Non-blocking）**：不会给每条消息增加延迟
|
||||
- **完整上下文（Complete context）**：可以访问完整的会话记录
|
||||
|
||||
## Related
|
||||
## 相关内容
|
||||
|
||||
- [The Longform Guide](https://x.com/affaanmustafa/status/2014040193557471352) - Section on continuous learning
|
||||
- `/learn` command - Manual pattern extraction mid-session
|
||||
- [长篇指南（The Longform Guide）](https://x.com/affaanmustafa/status/2014040193557471352) - 关于持续学习的部分
|
||||
- `/learn` 命令 - 在会话中手动提取模式
|
||||
|
||||
---
|
||||
|
||||
## Comparison Notes (Research: Jan 2025)
|
||||
## 对比笔记（研究：2025年1月）
|
||||
|
||||
### vs Homunculus (github.com/humanplane/homunculus)
|
||||
|
||||
Homunculus v2 takes a more sophisticated approach:
|
||||
Homunculus v2 采用了更复杂的方法：
|
||||
|
||||
| Feature | Our Approach | Homunculus v2 |
|
||||
| 特性（Feature） | 我们的方法（Our Approach） | Homunculus v2 |
|
||||
|---------|--------------|---------------|
|
||||
| Observation | Stop hook (end of session) | PreToolUse/PostToolUse hooks (100% reliable) |
|
||||
| Analysis | Main context | Background agent (Haiku) |
|
||||
| Granularity | Full skills | Atomic "instincts" |
|
||||
| Confidence | None | 0.3-0.9 weighted |
|
||||
| Evolution | Direct to skill | Instincts → cluster → skill/command/agent |
|
||||
| Sharing | None | Export/import instincts |
|
||||
| 观测（Observation） | 停止钩子（Stop hook，会话结束时） | PreToolUse/PostToolUse 钩子（100% 可靠） |
|
||||
| 分析（Analysis） | 主上下文（Main context） | 后台智能体（Background agent，Haiku） |
|
||||
| 粒度（Granularity） | 完整技能（Full skills） | 原子化的“本能（instincts）” |
|
||||
| 置信度（Confidence） | 无 | 0.3-0.9 加权 |
|
||||
| 演进（Evolution） | 直接转化为技能 | 本能（Instincts）→ 聚类（cluster）→ 技能/命令/智能体 |
|
||||
| 共享（Sharing） | 无 | 导出/导入本能 |
|
||||
|
||||
**Key insight from homunculus:**
|
||||
> "v1 relied on skills to observe. Skills are probabilistic—they fire ~50-80% of the time. v2 uses hooks for observation (100% reliable) and instincts as the atomic unit of learned behavior."
|
||||
**来自 homunculus 的关键洞察：**
|
||||
> “v1 依赖技能进行观测。技能是概率性的——它们的触发率约为 50-80%。v2 使用钩子进行观测（100% 可靠），并将本能（instincts）作为学习行为的原子单位。”
|
||||
|
||||
### Potential v2 Enhancements
|
||||
### 潜在的 v2 增强功能
|
||||
|
||||
1. **Instinct-based learning** - Smaller, atomic behaviors with confidence scoring
|
||||
2. **Background observer** - Haiku agent analyzing in parallel
|
||||
3. **Confidence decay** - Instincts lose confidence if contradicted
|
||||
4. **Domain tagging** - code-style, testing, git, debugging, etc.
|
||||
5. **Evolution path** - Cluster related instincts into skills/commands
|
||||
1. **基于本能的学习（Instinct-based learning）** - 带有置信度评分的小型原子化行为
|
||||
2. **后台观测器（Background observer）** - 并行分析的 Haiku 智能体
|
||||
3. **置信度衰减（Confidence decay）** - 如果出现矛盾，本能将失去置信度
|
||||
4. **领域标签（Domain tagging）** - 代码风格（code-style）、测试（testing）、git、调试（debugging）等
|
||||
5. **演进路径（Evolution path）** - 将相关的本能聚类为技能/命令
|
||||
|
||||
See: `/Users/affoon/Documents/tasks/12-continuous-learning-v2.md` for full spec.
|
||||
参见：`/Users/affoon/Documents/tasks/12-continuous-learning-v2.md` 以获取完整规范。
|
||||
|
||||
Reference in New Issue
Block a user