← 返回 Skills 市场
Self Improving Agent 3.0.6
作者
Raidan Pro
· GitHub ↗
· v1.0.0
· MIT-0
105
总下载
0
收藏
9
当前安装
1
版本数
在 OpenClaw 中安装
/install self-improving-agent-3-0-6
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 to do what it says (log learnings), but it asks you to create and modify files under ~/.openclaw, enable hooks, and (optionally) clone a third‑party GitHub repo. Before installing or enabling hooks: 1) Inspect the GitHub repo and any hook scripts for unexpected behavior; don't run them blindly. 2) Decide whether logs (.learnings/*, promoted files) could contain secrets or PII — redact or exclude sensitive data and add gitignore rules before promoting into a repo. 3) Be cautious with inter-session features (sessions_history, sessions_send) since they access other session transcripts. 4) If you want only ephemeral logging, avoid enabling persistent hooks or automatic promotion steps. If you want me to, I can list specific files/paths mentioned in the skill and sketch a safer installation checklist.
功能分析
Type: OpenClaw Skill
Name: self-improving-agent-3-0-6
Version: 1.0.0
The skill bundle implements a 'self-improvement' loop that instructs the agent to modify its own persistent instruction files (e.g., SOUL.md, AGENTS.md) based on session interactions, which creates a high risk for persistent prompt injection. Furthermore, SKILL.md references several shell scripts (scripts/activator.sh, scripts/error-detector.sh, and scripts/extract-skill.sh) that are not included in the provided files, preventing a full security review of the execution logic. The requested capabilities, including broad filesystem access and inter-session communication (sessions_history, sessions_send), are powerful and could be easily repurposed for unauthorized data access or persistence.
能力评估
Purpose & Capability
The name/description (self-improvement via logging) matches the instructions to append entries to .learnings and promote learnings to workspace files. This behavior is expected for a logging/self-improvement skill. However, the skill metadata declares no required config paths or install steps while the instructions assume an OpenClaw workspace (~/.openclaw/workspace) and hooks directory — a minor mismatch.
Instruction Scope
SKILL.md tells the agent/user to create and write to ~/.openclaw/workspace/.learnings and to copy/enable hooks under ~/.openclaw/hooks; it also references inter-session primitives (sessions_list, sessions_history, sessions_send, sessions_spawn). These actions imply reading/writing session transcripts and cross-session sharing which can expose sensitive conversation content and modify user config. The skill's instructions permit broad file writes and promotion of entries into project files (CLAUDE.md, AGENTS.md, .github/copilot-instructions.md), which could unintentionally commit secrets into a repo.
Install Mechanism
There is no automated install spec (instruction-only). The manual install suggestions include git clone from a GitHub repo (pskoett/pskoett-ai-skills). That is a typical pattern but cloning and copying hooks from an external repo pulls code/scripts from a third party and should be reviewed before running; no hashes or trusted release sources are provided.
Credentials
The skill declares no required environment variables or credentials, which is appropriate. However, the logging workflow encourages capturing runtime errors and conversation context; users should be aware that logs may include secrets, tokens, or PII unless the skill explicitly redacts them (it does not).
Persistence & Privilege
The instructions walk the user through enabling a hook (openclaw hooks enable self-improvement) and copying hook files into ~/.openclaw/hooks, which establishes persistent behavior that runs across sessions. The skill metadata does not declare this persistence or config path requirements. Persistent hooks that share learnings across sessions enlarge the blast radius if sensitive data is logged or if hooks execute code from an external repo.
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-improving-agent-3-0-6 - 安装完成后，直接呼叫该 Skill 的名称或使用
/self-improving-agent-3-0-6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-improving-agent skill for OpenClaw and generic agent environments:
- Provides a standardized workflow and markdown templates for logging errors, learnings, and feature requests.
- Outlines clear procedures for promoting important learnings to persistent project memory files (AGENTS.md, SOUL.md, TOOLS.md, etc).
- Offers detailed logging formats and metadata standards for learnings, errors, and feature requests.
- Includes both OpenClaw-specific and generic agent setup instructions.
- Describes session-to-session communication and optional hook integration for better workflow reminders.
元数据
常见问题
Self Improving Agent 3.0.6 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 105 次。
如何安装 Self Improving Agent 3.0.6？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install self-improving-agent-3-0-6」即可一键安装，无需额外配置。
Self Improving Agent 3.0.6 是免费的吗？
是的，Self Improving Agent 3.0.6 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Self Improving Agent 3.0.6 支持哪些平台？
Self Improving Agent 3.0.6 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 Improving Agent 3.0.6？
由 Raidan Pro（@raidan-ai）开发并维护，当前版本 v1.0.0。
推荐 Skills