據 AlphaMoat 統計,截至 2026-10,Ai Compound 1.0.1 累計下載 6,052 次、獲 5 個星標,按累計下載量在本次參與排名的 10,914 個 AI 技能中排名第 4,766 位,在「AI 智慧體」分類(1,632 個)中排名第 645 位。
列表預設按累計下載量排序,也可按星標排序;詳情頁排名按累計下載量計算。
| 平台 | 累計下載 | 星標 | 評論 |
|---|---|---|---|
| ClawHub ↗ | 3387 | 4 | - |
| SkillHub ↗ | 6052 | 5 | — |
各平台的計數口徑不同,統計時間不一致。
讓AI智慧體自動學習與提升。回顧會話,提取經驗,更新記憶檔案,持續積累知識。設定夜間審查迴圈,使智慧體每天更加聰明。
2026-09 起統計方法調整:同名技能按作者分別統計,並覆蓋全部技能。圖中虛線為調整前的月份,與之後的資料不直接比較,也不計算跨越調整的增長。
| 月份 | 累計下載 | 環比 | 星標 |
|---|---|---|---|
| 2026-10 | 6052 | — | 5 |
| 2026-09 | 3627 | — | 4 |
| 2026-08 | 3167 | +4.0% | 4 |
| 2026-07 | 3044 | — | 4 |
| 2026-05 | 2704 | — | 4 |
以下是作者釋出的原始檔案,按純文字顯示,本站未做修改或稽核;其中的連結不可點選,命令請核實後再執行。抓取於 2026-10-23。
--- name: compound-engineering description: Make your AI agent learn and improve automatically. Reviews sessions, extracts learnings, updates memory files, and compounds knowledge over time. Set up nightly review loops that make your agent smarter every day. version: 1.0.0 author: lxgicstudios keywords: compound, learning, memory, automation, agents, clawdbot, improvement, knowledge --- # Compound Engineering
Make your AI agent learn automatically. Extract learnings from sessions, update memory files, and compound knowledge over time.
**The idea**: Your agent reviews its own work, extracts patterns and lessons, and updates its instructions. Tomorrow's agent is smarter than today's.
---
## Quick Start
```bash
# Review last 24 hours and update memory
npx compound-engineering review
# Create hourly memory snapshot
npx compound-engineering snapshot
# Set up automated nightly review (cron)
npx compound-engineering setup-cron
```
---
## How It Works
### The Compound Loop
```
┌─────────────────────────────────────────┐
│ DAILY WORK │
│ Sessions, chats, tasks, decisions │
└────────────────┬────────────────────────┘
│
▼
┌─────────────────────────────────────────┐
│ NIGHTLY REVIEW (10:30 PM) │
│ • Scan all sessions from last 24h │
│ • Extract learnings and patterns │
│ • Update MEMORY.md and AGENTS.md │
│ • Commit and push changes │
└────────────────┬────────────────────────┘
│
▼
┌─────────────────────────────────────────┐
│ NEXT DAY │
│ Agent reads updated instructions │
│ Benefits from yesterday's learnings │
└─────────────────────────────────────────┘
```
### What Gets Extracted
- **Patterns**: Recurring approaches that worked
- **Gotchas**: Things that failed or caused issues
- **Preferences**: User preferences discovered
- **Decisions**: Key decisions and their reasoning
- **TODOs**: Unfinished items to remember
---
## Clawdbot Integration
### Automatic Hourly Memory
Add to your `HEARTBEAT.md`:
```markdown
# Hourly Memory Snapshot
Every hour, append a brief summary to memory/YYYY-MM-DD.md:
- What was accomplished
- Key decisions made
- Anything to remember
```
Or use cron:
```bash
# Add to clawdbot config or crontab
0 * * * * clawdbot cron run compound-hourly
```
### Nightly Review Job
Add this cron job to Clawdbot:
```json
{
"id": "compound-nightly",
"schedule": "30 22 * * *",
"text": "Review all sessions from the last 24 hours. For each session, extract: 1) Key learnings and patterns, 2) Mistakes or gotchas to avoid, 3) User preferences discovered, 4) Unfinished items. Update MEMORY.md with a summary. Update memory/YYYY-MM-DD.md with details. Commit changes to git."
}
```
---
## Manual Review Command
When you want to trigger a review manually:
```
Review the last 24 hours of work. Extract:
1. **Patterns that worked** - approaches to repeat
2. **Gotchas encountered** - things to avoid
3. **Preferences learned** - user likes/dislikes
4. **Key decisions** - and their reasoning
5. **Open items** - unfinished work
Update:
- MEMORY.md with significant long-term learnings
- memory/YYYY-MM-DD.md with today's details
- AGENTS.md if workflow changes needed
Commit changes with message "compound: daily review YYYY-MM-DD"
```
---
## Memory File Structure
### MEMORY.md (Long-term)
```markdown
# Long-Term Memory
## Patterns That Work
- When doing X, always Y first
- User prefers Z approach for...
## Gotchas to Avoid
- Don't do X without checking Y
- API Z has rate limit of...
## User Preferences
- Prefers concise responses
- Timezone: PST
- ...
## Project Context
- Main repo at /path/to/project
- Deploy process is...
```
### memory/YYYY-MM-DD.md (Daily)
```markdown
# 2026-01-28 (Tuesday)
## Sessions
- 09:00 - Built security audit tool
- 14:00 - Published 40 skills to MoltHub
## Decisions
- Chose to batch publish in parallel (5 sub-agents)
- Security tool covers 6 check categories
## Learnings
- ClawdHub publish can timeout, retry with new version
- npm publish hangs sometimes, may need to retry
## Open Items
- [ ] Finish remaining MoltHub uploads
- [ ] Set up analytics tracker
```
---
## Hourly Snapshots
For more granular memory, create hourly snapshots:
```bash
# Creates memory/YYYY-MM-DD-HH.md every hour
*/60 * * * * echo "## $(date +%H):00 Snapshot" >> ~/clawd/memory/$(date +%Y-%m-%d).md
```
Or have the agent do it via heartbeat by checking time and appending to daily file.
---
## The Compound Effect
**Week 1**: Agent knows basics
**Week 2**: Agent remembers your preferences
**Week 4**: Agent anticipates your needs
**Month 2**: Agent is an expert in your workflow
Knowledge compounds. Every session makes future sessions better.
---
## Setup Scripts
### Nightly Review (launchd - macOS)
```xml
<!-- ~/Library/LaunchAgents/com.clawdbot.compound-review.plist -->
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "...">
<plist version="1.0">
<dict>
<key>Label</key>
<string>com.clawdbot.compound-review</string>
<key>ProgramArguments</key>
<array>
<string>/opt/homebrew/bin/clawdbot</string>
<string>cron</string>
<string>run</string>
<string>compound-nightly</string>
</array>
<key>StartCalendarInterval</key>
<dict>
<key>Hour</key>
<integer>22</integer>
<key>Minute</key>
<integer>30</integer>
</dict>
</dict>
</plist>
```
### Hourly Memory (crontab)
```bash
# Add with: crontab -e
0 * * * * /opt/homebrew/bin/clawdbot cron run compound-hourly 2>&1 >> ~/clawd/logs/compound.log
```
---
## Best Practices
1. **Review before sleep** - Let the nightly job run when you're done for the day
2. **Don't over-extract** - Focus on significant learnings, not noise
3. **Prune regularly** - Remove outdated info from MEMORY.md monthly
4. **Git everything** - Memory files should be version controlled
5. **Trust the compound** - Effects are subtle at first, dramatic over time
---
Built by **LXGIC Studios** - [@lxgicstudios](https://x.com/lxgicstudios)
---
**Built by LXGIC Studios**
- GitHub: [github.com/lxgicstudios/ai-compound](https://github.com/lxgicstudios/ai-compound)
- Twitter: [@lxgicstudios](https://x.com/lxgicstudios)
將開發過程中的經驗、錯誤與修正記錄為 Markdown 檔案,供日後重用並提升為工作區記憶。
用自然語言描述需求,在本機與遠端技能市場尋找技能並取得推薦。
給 AI agent 用的瀏覽器自動化命令列技能。
將使用者明確糾正與自我檢視整理成分層記憶,並依使用情況升級或封存。
規劃 AI 代理主動跟進、工作記憶與回顧流程的工作區設定指南。
指導如何建立與更新有效 Claude 技能的寫作指南。
在安裝第三方技能之前,按清單審查來源、許可權與危險模式的技能。
把盛唐詩人李白的詩學觀、意象系統與口吻,蒸餾成 AI 可呼叫的角色包。
資料月份:2026-10 · 下載/星標數聚合自公開技能註冊庫(ClawHub、SkillHub),按月更新。口徑說明見資料方法。