reflect

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Reflect

复盘

Extract learnings from sessions and persist to skill files.

从会话中提炼经验并保存至技能文件。

When to Reflect

何时进行复盘

Run

/reflect

after sessions with:

Corrections - "actually use X", "no, do it this way"
Discoveries - patterns that worked well
Failures - approaches that didn't work

在以下类型的会话结束后运行

/reflect

：

修正内容 - “实际使用X”、“不对，应该这么做”
发现总结 - 有效的模式
失败教训 - 无效的方法

Usage

使用方法

/reflect              # Analyze current session, suggest skill updates
/reflect skill-name   # Target specific skill for updates

Note: This is manual-only. Run before ending sessions with learnings.

/reflect              # 分析当前会话，建议技能更新
/reflect skill-name   # 针对特定技能进行更新

注意： 此功能仅支持手动触发。请在结束有经验可提炼的会话前运行。

Why Manual-Only?

为何仅支持手动触发？

Auto-detection via Stop hooks doesn't work reliably:

Issue	Problem
#16227	Stop hook output is silent (not shown to user)
#10412	Plugin Stop hooks fail with exit code 2
#10875	Plugin JSON output not captured
#3656	Blocking functionality partially removed

Stop hooks fire but can't communicate back - making them useless for reminders. Other self-improving skill implementations (autoskill, reflect-skill) also use manual triggering for this reason.

通过Stop钩子进行自动检测并不可靠：

问题编号	具体问题
#16227	Stop钩子输出无提示（用户不可见）
#10412	插件Stop钩子执行失败，退出码为2
#10875	插件JSON输出未被捕获
#3656	阻塞功能被部分移除

Stop钩子会触发但无法回传信息——导致其无法用于提醒。其他自我改进技能实现（如autoskill、reflect-skill）也因此采用手动触发方式。

Process

流程

Source - Determine conversation source (see Data Sources below)
Analyze - Find corrections, successes, patterns
Classify - High/Medium/Low confidence learnings
Propose - Show suggested skill updates
Approve - Wait for user confirmation before writing

来源 - 确定对话来源（见下方数据源）
分析 - 找出修正内容、成功案例、有效模式
分类 - 将经验分为高/中/低置信度
提议 - 展示建议的技能更新内容
确认 - 等待用户确认后再写入文件

Data Sources

数据源

Try sources in order, use first available:

按顺序尝试以下来源，使用第一个可用的：

1. ccrecall.db (Full History)

1. ccrecall.db（完整历史）

If user has ccrecall + mcp-sqlite-tools:

sql

SELECT timestamp, role, content FROM messages
WHERE session_id = (SELECT MAX(session_id) FROM sessions)
ORDER BY timestamp DESC LIMIT 100;

如果用户安装了ccrecall + mcp-sqlite-tools：

sql

SELECT timestamp, role, content FROM messages
WHERE session_id = (SELECT MAX(session_id) FROM sessions)
ORDER BY timestamp DESC LIMIT 100;

2. In-Context (Current Session)

2. 上下文内（当前会话）

Fallback when ccrecall unavailable:

Analyze conversation visible in current context window
Limited to ~100k tokens of recent history
Still effective for single-session learnings

Note which mode is active:

[reflect] Using: ccrecall.db (full history)
-- or --
[reflect] Using: in-context (current session only)

当ccrecall不可用时的备选方案：

分析当前上下文窗口中可见的对话内容
限制为约10万token的近期历史内容
对单会话经验提炼仍然有效

注意当前激活的模式：

[reflect] 使用：ccrecall.db（完整历史）
-- or --
[reflect] 使用：上下文内（仅当前会话）

Destination Logic

目标路径逻辑

In repo with
```
.claude-plugin/
```
→ update skill in-place
Otherwise → prompt: project
```
.claude/skills/
```
or global
```
~/.claude/skills/
```

若在包含
```
.claude-plugin/
```
的仓库中 → 就地更新技能
否则 → 提示选择：项目目录下的
```
.claude/skills/
```
或全局目录
```
~/.claude/skills/
```

References

参考资料

analysis-patterns.md - Pattern detection rules

analysis-patterns.md - 模式检测规则