com um clique
system-metabolism
系统代谢·自动优化机制 — 每周体检、自动修剪、自我进化
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Menu
系统代谢·自动优化机制 — 每周体检、自动修剪、自我进化
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
Automate web browser interactions using natural language via CLI commands. Use when the user asks to browse websites, navigate web pages, extract data from websites, take screenshots, fill forms, click buttons, or interact with web applications.
Analyzes websites to generate multi-language content strategy, keyword research, and competitor analysis. Use when user provides a URL and needs industry analysis, SEO planning, or competitor research for content marketing in specific countries and languages.
Build and execute a content marketing strategy for a solopreneur business. Use when planning what content to create, deciding on content formats and channels, building a content calendar, measuring content performance, or systematizing content production. Covers audience research for content, content pillars, distribution strategy, repurposing workflows, and metrics. Trigger on "content strategy", "content marketing", "what content should I create", "content plan", "content calendar", "content ideas", "content distribution", "grow through content".
📊 Powerful ECharts-based data visualization skill optimized for Feishu (Lark) ecosystem. Supports 12+ chart types, 6+ data sources (Excel/CSV/Bitable/Sheet/Markdown), auto chart recommendation, auto analysis reports, generates high-definition PNG charts perfectly displayed in Feishu. No configuration required, works out of the box.
Analyze CSV/Excel files to extract insights, generate statistics, create charts, and produce summaries. Use when user wants to (1) upload or analyze spreadsheet data, (2) get insights from data files, (3) generate charts or visualizations, (4) calculate statistics or trends, (5) clean or transform data.
Knowledge graph operations via Graphiti API. Search facts, add episodes, and extract entities/relationships.
| name | system-metabolism |
| description | 系统代谢·自动优化机制 — 每周体检、自动修剪、自我进化 |
系统在日常运行中会自然积累:文件膨胀、指令冗余、任务权重下降、死指令遗留。需要一套定期体检与自动修剪的机制,让系统"越用越轻快",而非"越用越臃肿"。
┌────────────────────────────────────┐
│ 系统代谢(每周日 09:00) │
├────────────────────────────────────┤
│ ① 诊断层:采集8项健康指标 │
│ ② 判断层:识别需要优化的项目 │
│ ③ 执行层:自动修剪(安全项) │
│ ④ 报告层:推送给站长 + 记录日志 │
└────────────────────────────────────┘
使用以下命令采集指标:
# 1. 检查所有cron失败次数
openclaw cron list --include-disabled
# 2. 检查大文件
find ~/.openclaw/workspace/wiki -name "*.md" -exec wc -l {} + | sort -rn | head -15
find ~/.openclaw/workspace/memory -name "*.md" -exec wc -l {} + | sort -rn | head -15
# 3. 检查各任务指令长度(通过读取cron配置中的prompt)
| # | 指标 | 获取方式 | 阈值 | 自动处理 |
|---|---|---|---|---|
| 1 | Cron连续失败次数 | cron list | ≥3 | 标记+报修 |
| 2 | 文件行数 | wc -l | >300行 | 截断至最近50行(限append-only的日志文件) |
| 3 | 文件体积 | du -sh | >500KB | 标记为待优化 |
| 4 | 任务指令长度 | 读取cron payload | >800字 | 报告给站长 |
| 5 | 最近7天Token消耗 | 读取cron run stats | 平均>100k/次 | 标记为"肥胖任务" |
| 6 | 任务执行成功率 | 读取最近10次 | <70% | 标记+报修 |
| 7 | 死指令/不存在文件引用 | 指令中检查 | 发现即报告 | 自动清除指令 |
| 8 | 废弃任务 | 最近30天无成功 | 发现即报告 | 建议禁用 |
安全项(无需站长确认):
截断append-only日志文件(如 营销归因日志.md、案例库等)
exec tail -n 50 > 文件 保留最近50行清理死指令(如引用不存在的文件路径)
read / file 指令的文件路径exec test -f 验证存在性禁用连续失败≥7天的任务
cron update enabled=false需站长确认的(推送到飞书):
自动执行的变更记录到:
exec echo '| 2026-06-07 | 文件截断 | wiki/xxx.md | 从420行→50行 | ✅ 安全 |' >> ~/.openclaw/workspace/memory/topics/system-metabolism-log.md
推送到站长飞书(电影小镇群)的卡片结构:
## 📋 系统代谢周报 | 2026-06-07
### ✅ 已自动优化
- 截断 营销归因日志.md(997→50行)
- 禁用 xxx 任务(连续失败7天)
- 清理 2 条死指令
### ⚠️ 需您确认
- 竞品爆款拆解 指令精简方案(983字→建议精简至400字)
- 废弃任务:xxx(30天无成功,建议禁用)
### 📊 本周趋势
- 健康任务:32/35(↑2)
- 平均Token消耗:78k(↓22%)
- Token浪费:-58k/天
每次执行后,在 memory/topics/system-metabolism-log.md 追加一行,格式:
| 日期 | 操作 | 对象 | 效果 | 后续 |
|------|------|------|------|------|
| 6/7 | 截断 | 归因日志.md | 997→50行 | ✅ 已稳定 |
| 6/7 | 精简 | 营销效果归因 | token 290k→30k | ✅ 已稳定 |
累计足够日志后,可回溯分析:哪些优化最有效?哪些是白费功夫?