skillagi

记住会话中的错误。追加简短的学习内容，在复杂的任务之前进行回顾，促进重要的任务。

# Skillagi

Log mistakes and insights so you don't repeat them across sessions.

## When to Log

Add a learning when you:
- Fix a bug caused by a wrong assumption
- Discover a project-specific convention (package manager, config, naming)
- Get corrected by the user
- Find a workaround for a tool/platform limitation
- Waste time on an approach that doesn't work here

## Format

Append one line to `learnings.md` (in this skill's directory):

```
- [YYYY-MM-DD] topic: what happened → what to do instead
```

Keep it to one line. If you need two sentences, you're over-explaining.

## Examples

```
- [2026-02-21] pnpm not npm: Project uses pnpm workspaces → use `pnpm install`
- [2026-02-21] Docker M1: Base image has no ARM64 variant → add `--platform linux/amd64`
- [2026-02-21] test isolation: Shared DB state caused flaky tests → use transactions with rollback
```

## When to Review

Before starting a complex task, read `learnings.md` and apply anything relevant.
Don't review for trivial one-line changes.

## When to Clean Up

Periodically (or when the file exceeds ~50 entries):
- Delete entries that no longer apply (dependency upgraded, config changed)
- Merge duplicates into a single entry
- Promote critical entries that affect every session to a permanent location

## Promotion

Move persistent, high-impact learnings out of this file and into durable config:

- **OpenClaw**: Promote to workspace `MEMORY.md`, `TOOLS.md`, or `AGENTS.md`
- **Claude Code**: Promote to project `CLAUDE.md` or auto-memory (`~/.claude/projects/*/memory/`)

Once promoted, delete the entry from `learnings.md` to avoid duplication.

## Rules

- No IDs, statuses, priorities, or tags. Just the one-liner.
- No hooks or scripts. This is a passive reference file.
- Append-only during a session. Clean up between sessions.
- If unsure whether something is worth logging, skip it. Only log what would save future time.

---

## 中文说明

# Skillagi

记录错误和洞见，以便跨会话不再重蹈覆辙。

## 何时记录

在出现以下情况时添加一条学习记录：
- 修复了由错误假设导致的 bug
- 发现了项目特定的约定（包管理器、配置、命名）
- 被用户纠正
- 找到了应对工具/平台限制的变通办法
- 在某个在此处行不通的方法上浪费了时间

## 格式

向 `learnings.md`（位于本技能目录中）追加一行：

```
- [YYYY-MM-DD] topic: what happened → what to do instead
```

保持一行。如果你需要两句话，说明你解释得太多了。

## 示例

```
- [2026-02-21] pnpm not npm: Project uses pnpm workspaces → use `pnpm install`
- [2026-02-21] Docker M1: Base image has no ARM64 variant → add `--platform linux/amd64`
- [2026-02-21] test isolation: Shared DB state caused flaky tests → use transactions with rollback
```

## 何时回顾

在开始一项复杂任务之前，阅读 `learnings.md` 并应用任何相关内容。
对于琐碎的单行改动不必回顾。

## 何时清理

定期清理（或当文件超过约 50 条条目时）：
- 删除不再适用的条目（依赖已升级、配置已更改）
- 将重复项合并为单条条目
- 将影响每个会话的关键条目提升到一个永久位置

## 提升

将持久的、高影响的学习记录从此文件中移出，放入持久化配置：

- **OpenClaw**：提升到工作区 `MEMORY.md`、`TOOLS.md` 或 `AGENTS.md`
- **Claude Code**：提升到项目 `CLAUDE.md` 或自动记忆（`~/.claude/projects/*/memory/`）

提升后，从 `learnings.md` 中删除该条目以避免重复。

## 规则

- 不要 ID、状态、优先级或标签。只要这一行。
- 不要钩子或脚本。这是一个被动的参考文件。
- 会话期间仅追加。会话之间进行清理。
- 如果不确定某事是否值得记录，就跳过。只记录能节省未来时间的内容。