self-improving-agent

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Self-Improving Agent

自我改进Agent

Captures learnings, errors, and corrections to enable continuous improvement across sessions.

捕获学习内容、错误和纠正信息，实现跨会话的持续改进。

Activation Triggers

激活触发条件

This skill automatically activates when:

Trigger	What Gets Logged	Location
Command fails	Error type, context, recovery suggestion	`logs/failures_detailed.jsonl`
User corrects	"No, that's wrong...", "Actually..."	`logs/corrections.jsonl`
Missing capability	"Can you X?" where X isn't available	`logs/missing_capabilities.jsonl`
API/tool fails	Failure pattern, suggested fix	`logs/failures_detailed.jsonl`
Better approach found	Optimization learned	`logs/learnings.jsonl`

本技能在以下场景自动激活：

触发条件	记录内容	存储位置
命令执行失败	错误类型、上下文、恢复建议	`logs/failures_detailed.jsonl`
用户给出纠正	"No, that's wrong...", "Actually..."这类表述	`logs/corrections.jsonl`
能力缺失	用户提出当前不支持的功能请求，如"你能做X吗？"	`logs/missing_capabilities.jsonl`
API/工具调用失败	失败模式、建议修复方案	`logs/failures_detailed.jsonl`
发现更优方案	学习到的优化点	`logs/learnings.jsonl`

How It Works

工作原理

1. Automatic Logging (via hooks)

1. 自动日志记录（通过钩子实现）

PostToolUse → enhanced-failure-logger.js → logs failures with context
UserMessage → correction-detector.js → detects "wrong/actually/try again"
Response → capability-tracker.js → detects unfulfilled requests

PostToolUse → enhanced-failure-logger.js → logs failures with context
UserMessage → correction-detector.js → detects "wrong/actually/try again"
Response → capability-tracker.js → detects unfulfilled requests

2. Learning Aggregation

2. 学习聚合

All learnings flow to

logs/learnings.jsonl

json

{
  "timestamp": "2026-01-26T12:00:00Z",
  "type": "user_correction|tool_failure|missing_capability",
  "category": "factual_error|command_failed|web_browsing",
  "description": "what was learned",
  "source": "which detector"
}

所有学习内容都会流入

logs/learnings.jsonl

：

json

{
  "timestamp": "2026-01-26T12:00:00Z",
  "type": "user_correction|tool_failure|missing_capability",
  "category": "factual_error|command_failed|web_browsing",
  "description": "what was learned",
  "source": "which detector"
}

3. Session Start Review

3. 会话启动回顾

On each session start, recent learnings are shown:

=== Learning Review ===
[Learnings] 47 total entries
[Recent]
  • [tool_failure] Bash failed: timeout - WebFetch to external API...
  • [user_correction] User corrected: "No, use the other file..."
[Corrections] 12 user corrections logged
[Capability Gaps] Top requested:
  • send emails (5x)
  • browse web (3x)

每次会话启动时，会展示近期的学习内容：

=== Learning Review ===
[Learnings] 47 total entries
[Recent]
  • [tool_failure] Bash failed: timeout - WebFetch to external API...
  • [user_correction] User corrected: "No, use the other file..."
[Corrections] 12 user corrections logged
[Capability Gaps] Top requested:
  • send emails (5x)
  • browse web (3x)

4. QAVR Integration

4. QAVR集成

Successful learnings boost Q-values for related memories, improving future retrieval.

成功的学习内容会提升相关记忆的Q值，优化后续的检索效果。

Manual Commands

手动命令

Review Learnings

回顾学习内容

bash

undefined

bash

undefined

Show all learnings

cat ~/.claude/logs/learnings.jsonl | tail -20

Show corrections only

cat ~/.claude/logs/corrections.jsonl | jq -s 'group_by(.correction_type) | map({type: .[0].correction_type, count: length})'

Show capability gaps report

node ~/.claude/scripts/hooks/capability-tracker.js --report

undefined

node ~/.claude/scripts/hooks/capability-tracker.js --report

undefined

Test Detection

检测功能测试

bash

undefined

bash

undefined

Test correction detector

node ~/.claude/scripts/hooks/correction-detector.js

Test failure logger

node ~/.claude/scripts/hooks/enhanced-failure-logger.js

Test capability tracker

node ~/.claude/scripts/hooks/capability-tracker.js

undefined

node ~/.claude/scripts/hooks/capability-tracker.js

undefined

Learning Categories

学习分类

Correction Types

纠正类型

```
factual_error
```
- Wrong information provided
```
retry_request
```
- User asked to try again
```
misunderstanding
```
- Misinterpreted the request
```
failed_solution
```
- Solution didn't work

```
factual_error
```
- 提供了错误信息
```
retry_request
```
- 用户要求重试
```
misunderstanding
```
- 误解了用户请求
```
failed_solution
```
- 提供的解决方案无效

Failure Types

失败类型

```
permission_error
```
- Access denied
```
not_found
```
- File/resource missing
```
timeout
```
- Operation timed out
```
network_error
```
- Connection issues
```
syntax_error
```
- Invalid syntax
```
api_error
```
- External API failed
```
command_failed
```
- Shell command failed
```
agent_failed
```
- Subagent failed

```
permission_error
```
- 访问被拒绝
```
not_found
```
- 文件/资源不存在
```
timeout
```
- 操作超时
```
network_error
```
- 连接问题
```
syntax_error
```
- 语法无效
```
api_error
```
- 外部API调用失败
```
command_failed
```
- Shell命令执行失败
```
agent_failed
```
- 子Agent运行失败

Capability Categories

能力分类

```
web_browsing
```
- Internet access requests
```
image_processing
```
- Image/photo handling
```
communication
```
- Email/messaging
```
database_access
```
- SQL/database queries
```
external_api
```
- Third-party services
```
memory_persistence
```
- Long-term memory

```
web_browsing
```
- 互联网访问请求
```
image_processing
```
- 图片/照片处理
```
communication
```
- 邮件/消息发送
```
database_access
```
- SQL/数据库查询
```
external_api
```
- 第三方服务调用
```
memory_persistence
```
- 长期记忆存储

Configuration

配置

settings.json

, these hooks enable self-improvement:

json

{
  "hooks": {
    "SessionStart": [...],  // Reviews learnings
    "PostToolUse": [
      {"matcher": "Bash", "hooks": [{"command": "enhanced-failure-logger.js"}]},
      {"matcher": "Task", "hooks": [{"command": "enhanced-failure-logger.js"}]}
    ]
  }
}

在

settings.json

中，以下钩子会启用自我改进功能：

json

{
  "hooks": {
    "SessionStart": [...],  // Reviews learnings
    "PostToolUse": [
      {"matcher": "Bash", "hooks": [{"command": "enhanced-failure-logger.js"}]},
      {"matcher": "Task", "hooks": [{"command": "enhanced-failure-logger.js"}]}
    ]
  }
}

Benefits

优势

Learn from mistakes - Don't repeat the same errors
Understand user preferences - Track what corrections mean
Identify skill gaps - Know what features to build
Improve over time - QAVR ranking gets better with feedback
Context persistence - Learnings survive session restarts

从错误中学习 - 不会重复犯相同的错误
理解用户偏好 - 追踪用户纠正内容背后的偏好
识别能力缺口 - 明确需要开发的功能方向
持续优化 - QAVR排序会随着反馈不断优化
上下文持久化 - 学习内容不会随着会话重启丢失

Integration with Other Skills

与其他技能的集成

Skill	Integration
QAVR	Successful learnings boost memory Q-values
Memory Consolidation	Periodic cleanup of old learnings
Confidence Check	Review learnings before major tasks
IR-v2	Use learnings to inform pattern selection

技能	集成逻辑
QAVR	成功的学习内容会提升记忆Q值
Memory Consolidation	定期清理过时的学习内容
Confidence Check	执行重要任务前回顾学习内容
IR-v2	基于学习内容优化模式选择

Files

文件结构

~/.claude/
├── logs/
│   ├── learnings.jsonl           # All learnings
│   ├── corrections.jsonl         # User corrections
│   ├── failures_detailed.jsonl   # Enhanced failure logs
│   ├── missing_capabilities.jsonl # Capability requests
│   └── capability_gaps.json      # Aggregated gaps
└── scripts/hooks/
    ├── correction-detector.js
    ├── enhanced-failure-logger.js
    ├── capability-tracker.js
    └── session-start.js (reviews learnings)

~/.claude/
├── logs/
│   ├── learnings.jsonl           # All learnings
│   ├── corrections.jsonl         # User corrections
│   ├── failures_detailed.jsonl   # Enhanced failure logs
│   ├── missing_capabilities.jsonl # Capability requests
│   └── capability_gaps.json      # Aggregated gaps
└── scripts/hooks/
    ├── correction-detector.js
    ├── enhanced-failure-logger.js
    ├── capability-tracker.js
    └── session-start.js (reviews learnings)