← 返回 Skills 市场
Agent Self Improve
作者
yuyonghao-123
· GitHub ↗
· v0.1.0
· MIT-0
623
总下载
0
收藏
2
当前安装
1
版本数
在 OpenClaw 中安装
/install yuyonghao-agent-self-improve
功能描述
Analyzes agent performance and optimizes prompts, parameters, workflows, and strategies for continuous self-improvement.
安全使用建议
This skill appears internally consistent and low-risk: review and run its tests locally (npm test) before using it in production, and avoid passing sensitive secrets into the analyze/improve callbacks (e.g., do not pass raw credentials or private data as test inputs). Note the package description mentions "code self-rewriting" but no such behavior exists in the provided files — if you plan to use an auto-updating or self-modifying variant in future, audit any file-write or network code carefully. If you want extra caution, run the skill in an isolated environment or sandbox when first integrating it with a live agent.
功能分析
Type: OpenClaw Skill
Name: yuyonghao-agent-self-improve
Version: 0.1.0
The bundle provides a legitimate framework for performance profiling and strategy optimization of AI agents. It includes modules for measuring execution time, memory, and CPU usage (src/performance-analyzer.js) and tools for prompt and parameter tuning (src/strategy-optimizer.js). No evidence of data exfiltration, malicious execution, or unauthorized access was found, and the code aligns with the stated purpose in SKILL.md.
能力评估
Purpose & Capability
Name/description match the implemented modules: a PerformanceAnalyzer and a StrategyOptimizer for prompts, parameters, and workflows. One minor mismatch: package.json mentions "code self-rewriting" but the provided code does not implement any file-modifying or self-rewriting behavior — this appears to be an overstated description rather than hidden functionality.
Instruction Scope
SKILL.md only instructs npm install, shows creating SelfImprovementSystem and calling analyze/improve. The runtime example calls agent.process(input) indirectly (the analyze API expects a function), so the skill itself does not read files, env vars, or send data externally. Users should note that any data passed into analyze/improve comes from the host agent and could include secrets if the agent provides them.
Install Mechanism
No install spec in the registry and SKILL.md suggests running npm install for the package sources provided. package.json has no external dependencies and there are no downloads from untrusted URLs or extraction steps. Installing the package is low-risk.
Credentials
The skill declares no required environment variables, credentials, or config paths. The code only uses standard Node APIs (perf_hooks, process.memory/cpu) and no external tokens or secrets are requested.
Persistence & Privilege
The skill is not forced-always, is user-invocable, and does not modify other skills or system configuration. It does not request elevated or persistent privileges.
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 yuyonghao-agent-self-improve - 安装完成后,直接呼叫该 Skill 的名称或使用
/yuyonghao-agent-self-improve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
v0.1.0
Agent Self-Improve 0.1.0 initial release:
- Implements performance analyzer for execution time, resource usage, bottleneck identification, and hotspot detection.
- Introduces strategy optimizer, supporting prompt optimization, parameter tuning, workflow restructuring, and strategy selection.
- Provides straightforward API for analyzing agent performance and executing improvements.
- Includes installation and usage documentation.
元数据
常见问题
Agent Self Improve 是什么?
Analyzes agent performance and optimizes prompts, parameters, workflows, and strategies for continuous self-improvement. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 623 次。
如何安装 Agent Self Improve?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install yuyonghao-agent-self-improve」即可一键安装,无需额外配置。
Agent Self Improve 是免费的吗?
是的,Agent Self Improve 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Agent Self Improve 支持哪些平台?
Agent Self Improve 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 Agent Self Improve?
由 yuyonghao-123(@yuyonghao-123)开发并维护,当前版本 v0.1.0。
推荐 Skills