القائمة
Automatically fix, enhance, and optimize other skills when problems occur. Use when a skill fails or produces unexpected results. Prevents similar issues in the future.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | skill-auto-evolve |
| description | Automatically fix, enhance, and optimize other skills when problems occur. Use when a skill fails or produces unexpected results. Prevents similar issues in the future. |
这个 skill 用于自动修复、增强和优化其他 skill。当一个 skill 执行失败或产生意外结果时，自动分析问题并更新对应的 skill 文件。
当出现以下情况时自动触发：
首先确定问题类型：
| 错误类型 | 诊断方法 |
|---|---|
| 脚本超时 | 检查 timeout 配置，网络状态 |
| 依赖缺失 | 检查 node_modules，npm install |
| Cookie 过期 | 检查 cookies.json 的 expires 字段 |
| 选择器失效 | 检查目标网站 DOM 结构变化 |
| 参数错误 | 检查传入参数格式和内容 |
| 路径问题 | 检查相对/绝对路径 |
在 skill 目录下创建 issues/ 子目录，记录问题：
# issues/YYYY-MM-DD-issue-name.md
## 问题描述
[错误信息或现象]
## 诊断过程
[分析步骤]
## 根本原因
[根本原因]
## 修复方案
[解决方案]
## 验证结果
[修复后的测试结果]
根据问题类型更新 SKILL.md：
## 已知问题与解决方案
### 问题：[问题名称]
**症状**：[错误表现]
**原因**：[根本原因]
**解决**：[修复方法]
### 问题：[问题名称]
...
如果是脚本问题，直接修复脚本：
# 常见修复模式
# 1. 增加超时时间
timeout: 60000 # 从 30000 增加到 60000
# 2. 添加重试逻辑
async function retry(fn, times = 3) {
for (let i = 0; i < times; i++) {
try {
return await fn()
} catch (e) {
if (i === times - 1) throw e
await sleep(1000 * (i + 1))
}
}
}
# 3. 添加错误处理
try {
// 操作
} catch (error) {
console.error('详细错误:', error)
await page.screenshot({ path: 'debug-error.png' })
throw error
}
修复后运行测试验证：
# 重新执行失败的 skill
node [script].js [args]
# 检查输出
echo $? # 0 表示成功
// 修复方案：增加超时和重试
const browser = await puppeteer.launch({
headless: false, // 调试时设为 false
args: ['--no-sandbox', '--disable-setuid-sandbox'],
protocolTimeout: 120000, // 增加协议超时
})
await page.goto(url, {
waitUntil: 'networkidle2',
timeout: 60000, // 增加导航超时
})
// 检查 cookie 有效性
function checkCookieExpiry(cookies) {
const now = Date.now() / 1000
for (const cookie of cookies) {
if (cookie.expires && cookie.expires < now) {
console.warn(`Cookie ${cookie.name} 已过期`)
return false
}
}
return true
}
// 使用多个备选选择器
const selectors = [
'.new-selector',
'.old-selector',
'[data-testid="target"]'
]
for (const selector of selectors) {
try {
await page.waitForSelector(selector, { timeout: 5000 })
break
} catch (e) {
continue
}
}
修复完成后输出：
✅ Skill 已自动更新
📍 Skill: [skill-name]
🐛 问题: [问题描述]
🔧 修复: [修复内容]
📁 更新文件: [SKILL.md / script.js]
验证: [成功/失败]
/skill-auto-evolve [失败的skill名称] [错误信息]
例如：
/skill-auto-evolve zhihu-publish "Navigation timeout of 30000 ms exceeded"
Use when a project needs a standard .claude workspace initialized or repaired incrementally. This skill audits the existing .claude folder, adds missing files and directories without overwriting existing rules/skills/settings, and can optionally evolve the structure over time for the current project.
Generate Zhihu article in the style of 22-庖丁解牛专题页-知乎图文终版.md
Use when the user wants to search GitHub repositories by keyword, inspect ranked candidates, compare likely matches, and optionally clone the selected repository into a specified local directory. Good for prompts like “帮我搜 GitHub 上相关仓库”, “帮我找并下载一个相关仓库”, “搜索 GitHub 上的项目然后拉到 reference”, or “按关键词找仓库并 clone 到指定位置”.
Use when the task is to build or maintain a repo-local 热门话题/AI日报专项, aggregate multiple public hot-topic sources such as AIbase Daily or RSS feeds, normalize them into a reusable snapshot, route the findings into the best existing site hotspot pages, and keep the hot-topic hub updated over time.
Use when the task is to take a user-provided keyword, search the web for relevant articles, summarize and cluster the findings, choose the most appropriate existing `site/` topic or hotspot page in this project, and draft a publish-ready Markdown article with explicit image prompt placeholders or Mermaid notes.
Use when a task should advance through repeated plan-driven loop cycles for an engineering project, paper, or long-running repo change. loloop wraps the official /loop idea with the repo's loop-in-loop method: read the current plan, inspect prior evolution notes, execute the current iteration, write a new evolution record, draft the next plan when a version is complete, and prepare the next /loop handoff.