experiment-notes

Track, search, and learn from experiments. Automatic logging of trial-and-error, success/failure patterns, and distilled lessons. Prevents repeating mistakes.

3,891 stars

Best use case

experiment-notes is best used when you need a repeatable AI agent workflow instead of a one-off prompt.

Track, search, and learn from experiments. Automatic logging of trial-and-error, success/failure patterns, and distilled lessons. Prevents repeating mistakes.

Teams using experiment-notes should expect a more consistent output, faster repeated execution, less prompt rewriting.

When to use this skill

  • You want a reusable workflow that can be run more than once with consistent structure.

When not to use this skill

  • You only need a quick one-off answer and do not need a reusable workflow.
  • You cannot install or maintain the underlying files, dependencies, or repository context.

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/experiment-notes/SKILL.md --create-dirs "https://raw.githubusercontent.com/openclaw/skills/main/skills/besty0121/experiment-notes/SKILL.md"

Manual Installation

  1. Download SKILL.md from GitHub
  2. Place it in .claude/skills/experiment-notes/SKILL.md inside your project
  3. Restart your AI agent — it will auto-discover the skill

How experiment-notes Compares

Feature / Agentexperiment-notesStandard Approach
Platform SupportNot specifiedLimited / Varies
Context Awareness High Baseline
Installation ComplexityUnknownN/A

Frequently Asked Questions

What does this skill do?

Track, search, and learn from experiments. Automatic logging of trial-and-error, success/failure patterns, and distilled lessons. Prevents repeating mistakes.

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

Related Guides

SKILL.md Source

# 实验笔记系统 / Experiment Notes

让 Agent 从经验中学习,不再重复犯错。

Every agent fails. Smart agents remember why.

## 核心理念

普通 Agent:
```
试了 → 失败了 → 忘了 → 下次又试 → 又失败了
```

用了实验笔记的 Agent:
```
试了 → 失败了 → 记下来了 → 下次查一下 → 直接用对的方案
```

## 安装

skill 安装后,数据存储在 `~/.openclaw/memory/experiments/`:
- `experiments.jsonl` — 所有实验记录
- `lessons.jsonl` — 提炼的最佳实践

首次使用会自动创建目录。

## 命令参考

所有命令通过 `python <skill_dir>/scripts/expnote.py` 调用。

### 记录实验

```bash
python expnote.py log \
  --task "从 Docker Hub 拉取镜像" \
  --outcome failure \
  --tags docker,network \
  --cmd "docker pull nginx" \
  --error "Error response from daemon: Get https://registry-1.docker.io/v2/: net/http: request canceled" \
  --fix "配置 Docker 镜像加速器" \
  --lesson "国内环境拉 Docker 镜像大概率超时,先配加速器"
```

**outcome 取值:**
- `success` — 成功了 ✅
- `failure` — 失败了 ❌
- `partial` — 部分成功 ⚠️

### 搜索历史

```bash
# 全文搜索
python expnote.py search "docker"

# 限制结果数
python expnote.py search "clawhub publish" --limit 5
```

### 查找类似任务

在开始新任务前,先查查之前有没有做过类似的:

```bash
python expnote.py similar "发布 skill 到 ClawHub"
```

### 查看经验教训

```bash
# 所有教训
python expnote.py lessons

# 按标签过滤
python expnote.py lessons --tag docker
```

### 统计概览

```bash
python expnote.py stats
```

显示:总实验数、成功率/失败率、热门标签、最近活动。

### 提炼最佳实践

从多次实验中总结通用经验:

```bash
python expnote.py distill \
  --tags clawhub,publish \
  --lesson "clawhub publish 必须显式指定 --version,不支持自动递增"
```

## 给 Agent 的使用指南

在你的 `AGENTS.md` 中添加以下规则:

```markdown
## 实验笔记

每次尝试新操作时:

1. **之前**:运行 `expnote.py similar "任务描述"`,看看有没有历史经验
2. **之后**:无论成功失败,运行 `expnote.py log` 记录结果
3. **定期**:review 实验记录,用 `expnote.py distill` 提炼通用教训

记录格式:
- task: 简短描述(动词开头)
- outcome: success / failure / partial
- tags: 相关技术/工具(逗号分隔)
- cmd: 实际执行的命令
- error: 报错信息(失败时)
- fix: 怎么修复的
- lesson: 一句话总结(最重要!)
```

## 数据格式

每条实验记录(experiments.jsonl 中的一行):

```json
{
  "id": "exp_a1b2c3d4e5f6",
  "timestamp": "2026-04-02T08:58:00+08:00",
  "task": "发布 ClawHub skill",
  "outcome": "partial",
  "attempt": {
    "command": "clawhub publish ./skill --no-input",
    "description": "发布小红书浏览器 skill"
  },
  "result": {
    "error": "Error: --version must be valid semver",
    "fix": "加上 --version 1.1.0"
  },
  "tags": ["clawhub", "publish", "skill"],
  "lesson": "clawhub publish 必须显式指定 --version"
}
```

## 目录结构

```
experiment-notes/
├── SKILL.md              # 本文件
├── scripts/
│   └── expnote.py        # CLI 工具
└── templates/
    └── agents-snippet.md  # AGENTS.md 集成片段
```

## 设计理念

- **JSONL 格式**:追加写入,不怕并发,方便 grep
- **纯文件存储**:不需要数据库,不需要服务端
- **标签系统**:多维分类,支持按领域过滤
- **提炼机制**:不只是记录,还要总结成可复用的知识

## 局限性

- 搜索是关键词匹配,不是语义搜索(够用但不完美)
- 跨 agent 共享需要手动同步文件
- 没有自动清理机制(可以手动删除旧记录)

Related Skills

Meeting Mastery — AI Meeting Prep, Notes & Follow-Up Engine

3891
from openclaw/skills

You are an elite meeting preparation and follow-up agent. You ensure every meeting is high-value — thoroughly prepared beforehand, cleanly documented during, and actioned after.

Workflow & Productivity

obsidian-notes

3891
from openclaw/skills

Work with Obsidian vaults (plain Markdown notes) and automate via obsidian-cli.

Workflow & Productivity

ml-experiment-tracker

3891
from openclaw/skills

Plan reproducible ML experiment runs with explicit parameters, metrics, and artifacts. Use before model training to standardize tracking-ready experiment definitions.

Data & Research

Notes (笔记)

3891
from openclaw/skills

> Prerequisites: see root `../SKILL.md` for setup, credentials, and `ima_api()` helper.

tasknotes

3891
from openclaw/skills

Manage tasks in Obsidian via TaskNotes plugin API. Use when user wants to create tasks, list tasks, query by status or project, update task status, delete tasks, or check what they need to do.

ai-notes-of-video

3891
from openclaw/skills

The video AI notes tool is provided by Baidu. Based on the video download address provided by the user, it downloads and parses the video, and finally generates AI notes corresponding to the video (a total of three types of notes can be generated: document notes, outline notes, and image-text notes).

experiment-designer

3891
from openclaw/skills

Use when planning product experiments, writing testable hypotheses, estimating sample size, prioritizing tests, or interpreting A/B outcomes with practical statistical rigor.

lead-qualification-notes

3891
from openclaw/skills

把销售线索信息整理成资格判断、关键问题、风险项与下一步建议。;use for sales, lead-qualification, crm workflows;do not use for 虚构客户预算, 替代 CRM 主数据.

36kr-ainotes

3891
from openclaw/skills

获取36氪AI测评的每日最新测评笔记内容. Use when the user asks about 36kr AI测评, AI测评笔记, 测评笔记, ainotes, 36kr测评, 查AI测评, 最新AI测评笔记, 今日AI测评, 36kr ainotes, 看测评, 查看测评内容, 36kr AI产品测评, 产品测评笔记, 今天有什么测评, 有哪些AI产品被测评了, 36aidianping测评笔记, 36kr ai notes.

Release Notes Generator

3891
from openclaw/skills

Generate clear, professional release notes from git commits, PRs, or a plain list of changes. Outputs formatted changelogs your users actually want to read.

gitcode-release-notes

3880
from openclaw/skills

Generate release notes for GitCode repositories from commits (by tag range or since-date), grouped as feat/fix/docs/other, output Markdown for Release pages. 按 tag 区间或日期拉取提交并生成版本发布公告 Markdown。Python 3.7+ standard library only.

---

3891
from openclaw/skills

name: article-factory-wechat

Content & Documentation