← Back to Skills Marketplace
110
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install self-improving-x
Description
捕获学习成果、错误和纠正，以实现持续改进，当出现以下情况，使用本技能： (1) 命令或操作意外失败， (2) 用户纠正 '不，那是错的...'，'实际上...'）， (3) 用户请求不存在的功能， (4) 外部 API 或工具失败， (5) 意识到其知识已过时或不正确， (6) 发现更好的方法处理重复任务。在执行...
Usage Guidance
概要建议和注意项：
- 这个技能只是文档/模板驱动，会在你的工作区创建和修改 .learnings/LEARNINGS.md、ERRORS.md、FEATURE_REQUESTS.md 以及可能的 AGENTS.md/CLAUDE.md 等项目文件；它不请求任何凭据或安装二进制。
- 在允许技能运行并自动写入仓库之前：备份或使用受控的测试工作区，审阅并确认将被添加的条目格式与内容，避免在受限或敏感代码库中直接启用自动写入。
- 技能会读取本地文件（如 .learnings/*、memory/*、项目文档）并运行简单 shell grep 示例，这是为了去重与回顾历史——如果不希望模型访问某些文件，请将这些路径排除或在受限环境中运行。
- README 中提到的 npx 安装（指向外部 GitHub）与本包的 instruction-only 形式是分开的：如果你用 npx 拉取外部仓库，务必审查该远端仓库的内容和脚本。
- 总体上该技能内部一致且用途明确：如果你的目标是让 agent 在项目中捕获并改进学习记录，它是合适的；若担心自动修改仓库，改为手动触发并在 PR 中审查更安全。
Capability Analysis
Type: OpenClaw Skill
Name: self-improving-x
Version: 1.0.0
The bundle is a documentation framework designed to help an AI agent track its own performance, errors, and user feedback within a local workspace. It uses Markdown files (SKILL.md, LEARNINGS.md, ERRORS.md) to instruct the agent to log 'learnings' and 'feature requests' into a `.learnings/` directory. There are no executable scripts, network exfiltration attempts, or malicious prompt injections; the logic is entirely focused on local workspace organization and workflow optimization.
Capability Assessment
Purpose & Capability
名称和描述为“自我改进/记录学习与错误”，所需内容是创建并维护 .learnings/ 下的 Markdown 日志并将重要条目提升到项目文档；技能没有要求与其用途不相关的凭据、二进制或安装，整体匹配声明目的。
Instruction Scope
SKILL.md 明确指示在工作区中创建/追加/更新多个markdown文件（.learnings/LEARNINGS.md、ERRORS.md、FEATURE_REQUESTS.md）、可能修改 AGENTS.md、SOUL.md、TOOLS.md、CLAUDE.md 等项目文档，并运行本地 grep 命令进行查找。这些都是为“记录与提升学习”合理且预期的操作，但会导致读写和修改项目文件；如果您不希望自动更改仓库文件，应在使用前备份或限制该技能的写入权限。
Install Mechanism
instruction-only（无安装脚本、无代码文件需要执行），因此没有下载或在本地写入可执行代码的高风险安装流程。README 建议一个 npx 安装命令（指向 GitHub），但技能包本身为文档模板，无隐蔽二进制或远程可执行载荷。
Credentials
不需要环境变量、凭据或配置路径。SKILL.md 只引用本地工作区文件与其它技能输出（例如 simplify-and-harden 的候选项），这些引用与记录/提升学习的目标一致，没有要求不相关的敏感凭据。
Persistence & Privilege
技能未设置 always:true，并允许正常的模型调用（默认）。尽管如此，运行时指示会修改工作区内的文件（创建/追加/提升条目至 AGENTS.md、CLAUDE.md 等），这等同于在项目中持久写入。该行为对技能目的合理，但在敏感或受保护的代码库中应谨慎授权写入权限。
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-improving-x - After installation, invoke the skill by name or use
/self-improving-x - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
**Major overhaul: Simplifies and restructures self-improving workflow, focusing on separate, structured markdown logs for learning, errors, and feature requests.**
- Replaces large, complex workspace memory and heartbeat files with modular `.learnings/` markdown files: `LEARNINGS.md`, `ERRORS.md`, `FEATURE_REQUESTS.md`.
- Removes previous architecture (`memory.md`, `heartbeat-state.md`, `corrections.md`, etc.) and many supporting rules/templates; instead, provides clear logging templates and file sample usage.
- Adds ready-to-copy asset templates and real-world examples for error, learning, and feature request logs.
- Establishes a repeatable process for promoting widely-applicable learnings to core workflow/config files (e.g., `AGENTS.md`, `SOUL.md`, `TOOLS.md`).
- Clarifies when and how to log, review, promote, and extract knowledge as reusable project skills.
- Reduces setup complexity, modularizes logs, and enables easier review, prioritization, and automation.
Metadata
Frequently Asked Questions
What is self-improving-x (support multi-agent)?
捕获学习成果、错误和纠正，以实现持续改进，当出现以下情况，使用本技能： (1) 命令或操作意外失败， (2) 用户纠正 '不，那是错的...'，'实际上...'）， (3) 用户请求不存在的功能， (4) 外部 API 或工具失败， (5) 意识到其知识已过时或不正确， (6) 发现更好的方法处理重复任务。在执行... It is an AI Agent Skill for Claude Code / OpenClaw, with 110 downloads so far.
How do I install self-improving-x (support multi-agent)?
Run "/install self-improving-x" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is self-improving-x (support multi-agent) free?
Yes, self-improving-x (support multi-agent) is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does self-improving-x (support multi-agent) support?
self-improving-x (support multi-agent) is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created self-improving-x (support multi-agent)?
It is built and maintained by seepine (@seepine); the current version is v1.0.0.
More Skills