← 返回 Skills 市场
Self Improvement Ai
作者
Fardeen Karim
· GitHub ↗
· v1.0.0
· MIT-0
142
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install self-improvement-ai
功能描述
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
安全使用建议
This skill appears coherent and low-risk for its stated goal, but before installing or enabling hooks: 1) Verify provenance (SKILL.md references GitHub but registry/homepage are blank) and trust the author. 2) Inspect the three shell scripts (activator.sh, error-detector.sh, extract-skill.sh) and hook handler files yourself; they are small and readable. 3) Note that enabling the hook causes those scripts/handlers to run with the agent's permissions and the error-detector reads CLAUDE_TOOL_OUTPUT (platform-provided tool output) — acceptable for local error detection but avoid enabling in multi-tenant or untrusted environments without review. 4) If you run extract-skill.sh, run in a safe working directory (it prevents absolute paths and '..' but will create files under the current path). 5) If unsure, enable only the activator (reminder) and skip PostToolUse/error detection until you've validated behavior.
功能分析
Type: OpenClaw Skill
Name: self-improvement-ai
Version: 1.0.0
The 'self-improvement-ai' skill bundle is a legitimate tool designed to help AI agents track errors, user corrections, and feature requests in a local `.learnings/` directory. It includes utility scripts like `scripts/error-detector.sh` to identify command failures and `scripts/extract-skill.sh` to scaffold new skills from logged insights, the latter of which includes basic path traversal protections. The OpenClaw hooks (`hooks/openclaw/handler.js`) and markdown instructions (`SKILL.md`) are focused on maintaining a project-specific memory and improving agent behavior without any evidence of data exfiltration, malicious execution, or unauthorized persistence.
能力评估
Purpose & Capability
The skill's name/description (capture learnings, errors, promote learnings) matches the provided files: documentation, templates, a hook to inject reminders, and helper scripts to detect errors and scaffold extracted skills. The presence of OpenClaw/OpenClaw-hook handlers, activator/error-detector scripts, and extract-skill.sh is appropriate for this purpose. Note: the package metadata lacks a homepage/source URL even though SKILL.md refers to GitHub repos; that is a provenance gap to verify but does not contradict the skill's functionality.
Instruction Scope
SKILL.md instructs creating .learnings/, copying/enabling an OpenClaw hook, and optionally running included scripts. The instructions stay within the stated domain (logging, reminding, extracting skills). The error detector reads the platform-provided CLAUDE_TOOL_OUTPUT variable to detect command failures — expected for a PostToolUse hook. Nothing in the instructions directs data to external endpoints or asks the agent to read unrelated secrets or system-wide configurations.
Install Mechanism
There is no automatic install/download step; it's instruction-only with local scripts and hooks included. No network downloads or archive extracts are performed by the skill itself. The extract-skill.sh writes files under a relative ./skills directory (it blocks absolute paths and '..'), which is reasonable for a scaffolding helper — review before running.
Credentials
The skill declares no required environment variables or credentials. Runtime scripts reference CLAUDE_TOOL_OUTPUT (a platform-provided variable) and standard shell behavior; this is proportionate to detecting tool errors. No unrelated secrets or external-service keys are requested.
Persistence & Privilege
always is false and the skill is user-invocable. Hook enablement is explicitly opt-in (you must copy/enable the hook). The hook injects a virtual reminder at bootstrap and scripts run with the same permissions as the agent when invoked — expected behavior for a hook-based reminder system. The skill does not modify other skills' configurations or demand permanent elevated presence.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install self-improvement-ai - 安装完成后，直接呼叫该 Skill 的名称或使用
/self-improvement-ai触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
Initial release of the self-improvement skill for continuous learning and error logging.
- Enables structured logging of errors, user corrections, feature requests, knowledge gaps, and best practices to markdown files.
- Supports both OpenClaw and generic agent setups, with detailed instructions for file setup and promotion to project memory.
- Introduces standardized formats for learnings, errors, and feature requests, including metadata for tracking and resolution.
- Provides workflow to review and promote broadly applicable learnings to core project files (AGENTS.md, SOUL.md, TOOLS.md, CLAUDE.md).
- Includes OpenClaw integration with hooks and workspace file structure for seamless adoption and inter-session communication.
元数据
常见问题
Self Improvement Ai 是什么？
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 142 次。
如何安装 Self Improvement Ai？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install self-improvement-ai」即可一键安装，无需额外配置。
Self Improvement Ai 是免费的吗？
是的，Self Improvement Ai 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Self Improvement Ai 支持哪些平台？
Self Improvement Ai 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Self Improvement Ai？
由 Fardeen Karim（@fardeenkarim）开发并维护，当前版本 v1.0.0。
推荐 Skills