← 返回 Skills 市场
112
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install self-improving-agent-tmp
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
What to check before installing or enabling:
- Verify the origin: the SKILL.md references installing 'self-improving-agent' but the registry entry is 'self-improving-agent-tmp' and _meta.json shows a different ownerId — confirm you trust the exact package/source before copying it into ~/.openclaw or enabling hooks.
- Review the scripts/hooks locally: activator.sh, error-detector.sh, and the hook handler (handler.js/ts) are small and readable; enable them only if you understand they will run in your sessions. They run with the same permissions as the agent and can execute if you add them to your hook configuration.
- Prefer project-level (not global) hook configuration or use matcher filters to limit when hooks run; avoid enabling user-level/global hooks unless you want reminders across all sessions.
- Test in a safe workspace first: use the extract-skill.sh --dry-run and test hooks in a sandbox workspace before deploying to important projects or enabling for all sessions.
- If you plan to let hooks run automatically, ensure scripts are executable and consider auditing them for any modifications. If anything looks unexpected (network calls, writes to unrelated system paths, or credential references), do not enable the hook.
Confidence notes: assessment is medium because the package contains multiple files and small metadata/name mismatches. Those mismatches are not proof of malice but do warrant verification of the source before use.
功能分析
Type: OpenClaw Skill
Name: self-improving-agent-tmp
Version: 1.0.0
The 'self-improving-agent' skill bundle is a legitimate framework designed to help AI agents log errors, user corrections, and insights for continuous improvement. It utilizes OpenClaw hooks (hooks/openclaw/handler.js) to inject reminders into the agent's context and provides utility scripts (scripts/extract-skill.sh) for scaffolding new skills based on learned patterns. The behavior is transparent, well-documented, and lacks any indicators of malicious intent such as data exfiltration or unauthorized remote execution.
能力评估
Purpose & Capability
The skill's documented purpose (log learnings/errors and optionally inject reminders) matches the included files: logging templates, hook handlers, and small helper scripts. However, there are minor inconsistencies in naming/metadata: registry slug/name include '-tmp' while install instructions and examples reference 'self-improving-agent' (no '-tmp'), and _meta.json lists a different ownerId than the registry metadata. These mismatches suggest repackaging or an upstream copy and deserve verification before install.
Instruction Scope
SKILL.md instructs only to create .learnings/, copy templates, and optionally install/enable hooks; hooks inject a virtual reminder and scripts detect command output errors or scaffold new skills. The instructions do not request unrelated system files or credentials. They do direct optionally copying files into user home (~/.openclaw) and enabling hooks which will cause the provided scripts/handlers to run in future sessions (expected for this purpose).
Install Mechanism
No automatic install spec is provided (instruction-only skill). The included scripts and hook handlers are local files (no remote downloads). The extract-skill.sh has path checks to avoid absolute/.. writes. Overall install risk is low because installation is manual and opt-in.
Credentials
The skill does not declare required environment variables or credentials. The error-detector.sh reads CLAUDE_TOOL_OUTPUT (a hook-provided tool output variable) which is appropriate for a hook that examines command results. No other secret or unrelated service credentials are requested.
Persistence & Privilege
always:false and hooks/scripts are opt-in; enabling hooks requires explicit user action. The hook injects only a virtual bootstrap file and the scripts write only to user-scoped paths when run (e.g., ./skills or ~/.openclaw). The skill does not modify other skills' configs or request permanent elevated 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 self-improving-agent-tmp - 安装完成后，直接呼叫该 Skill 的名称或使用
/self-improving-agent-tmp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
Self-Improvement Skill initial release for OpenClaw and generic agent frameworks.
- Provides structured workflows and markdown templates to log errors, corrections, learnings, and feature requests for continuous improvement.
- Details specific file conventions: `.learnings/LEARNINGS.md`, `ERRORS.md`, and `FEATURE_REQUESTS.md`.
- Outlines promotion criteria for important learnings to project-wide memory files.
- Includes setup instructions for OpenClaw integration and generic agents.
- Describes standardized logging formats, metadata, and workflow for handling and resolving entries.
- Supports multi-session learning sharing within OpenClaw.
元数据
常见问题
Self Improving Agent.Tmp 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 112 次。
如何安装 Self Improving Agent.Tmp？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install self-improving-agent-tmp」即可一键安装，无需额外配置。
Self Improving Agent.Tmp 是免费的吗？
是的，Self Improving Agent.Tmp 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Self Improving Agent.Tmp 支持哪些平台？
Self Improving Agent.Tmp 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.Tmp？
由 laner（@kelaner）开发并维护，当前版本 v1.0.0。
推荐 Skills