openclaw-skills/self-improving-agent/SKILL.md
Use when building or operating agents that should learn from prior runs, capture durable feedback, update behavior safely, and improve task performance across iterations.
npx skillsauth add seaworld008/commonly-used-high-value-skills self-improving-agentInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
用于需要长期运行、持续学习和策略改进的 Agent 工作流。Self Improving Agent 核心逻辑在于“闭环进化”:它不仅执行任务,还会观察执行结果、收集反馈,并据此修改自身的 Prompts、工具调用链或决策权重,从而在下一次类似任务中表现得更好。
npx clawhub@latest install self-improving-agent
System Prompt 为更简洁的风格。MEMORY.md 知识过时或发生冲突时,Agent 主动发起“知识清洗”任务。self-improving-agent 进行根因分析。cron 技能每周执行一次自我诊断(Self-Diagnostics)。process 技能运行长任务时,实时重定向 stderr 到审计文件。memory_search 查找过去 10 次同类任务的成败记录。edit 工具修改自身的 SKILL.md 正文或 System Prompt 模板。MEMORY.md 中的“禁忌法则”和“成功套路”。MEMORY.md 记录 diff 变更记录。mcp_call 将这些知识同步到共享知识库。### 原始任务执行记录 (Execution Log)
- **任务**: [生成 React 登录页面]
- **结果**: [失败 - 样式丢失]
- **反馈**: [用户反馈称样式未正确引入 Tailwind]
### 根因分析 (Root Cause)
- **诊断**: 原始 Prompt 中未显式包含 `tailwind.config.js` 的引用建议。
- **关联记忆**: 3 天前在项目 B 中也遇到了同样的问题。
### 优化决策 (Improvement Decision)
- **行动**: 更新 `frontend-design` 技能的 `Core Capabilities` 模板。
- **变更点**: 强制要求在所有 React 渲染任务中包含 `style_framework_check` 步骤。
### 下一步验证 (Verification)
- [ ] 运行 `test-tailwind-render` 命令。
- [ ] 等待用户下一次 React 相关指令。
// 示例:基于用户反馈动态调整说话风格
if (userFeedback.includes("concise")) {
await mcp_call({
name: 'update_system_prompt',
arguments: {
modifier: "Always provide answers in bullet points, max 3 items."
}
});
}
self-improving-agent 逻辑生成重构方案,并在 PR 中提出“这是我根据之前的代码模式总结出的通用组件”。SKILL.md 可能覆盖用户的手工配置,需保留 undo 机制。MEMORY.md 中有迹可循。development
Enumerating failure modes via pre-mortem analysis. Systematically identifies failure scenarios for plans, designs, and features, scoring them with RPN/AP. Does not write code.
testing
Orchestrating specialist AI agent teams as a meta-coordinator. Decomposes requests into minimum viable chains, spawns each as an independent session in AUTORUN modes, and drives to final output. Use when a task spans multiple specialist domains, requires parallel agent execution, or needs hub-and-spoke routing across the skill ecosystem.
development
Converting document formats (Markdown/Word/Excel/PDF/HTML). Converts specs from Scribe and reports from Harvest into distributable formats; generates reusable conversion scripts. Use when converting documents, building accessibility-compliant PDFs, or creating Pandoc/LibreOffice pipelines.
testing
Curating cross-agent knowledge and guarding institutional memory. Extracts patterns from agent journals into METAPATTERNS.md, detects knowledge decay, propagates best practices, prevents organizational forgetting. Use when consolidating cross-agent insights, curating memory, or auditing knowledge decay.