← 返回 Skills 市场
271
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install wangkang-skill
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 implements what it claims: simple reminders, error detection (from CLAUDE_TOOL_OUTPUT), and templates for logging learnings. Before enabling the hook globally, consider: 1) review .learnings entries before promoting them to workspace files (SOUL.md/AGENTS.md/TOOLS.md) because those files become part of the agent's context and can change behavior; 2) enable hooks at the project/user level (not system-wide) if you want to limit scope; 3) scripts are local and do not exfiltrate secrets, but check file permissions and inspect the files yourself if you have strict security requirements; 4) use extract-skill.sh --dry-run to see what it will create before writing files.
功能分析
Type: OpenClaw Skill
Name: wangkang-skill
Version: 1.0.1
The skill bundle implements a 'self-improvement' framework designed to help OpenClaw agents log errors, corrections, and best practices to markdown files (e.g., .learnings/LEARNINGS.md). The included shell scripts (activator.sh, error-detector.sh, and extract-skill.sh) are utility tools for triggering reminders and scaffolding new skills; notably, extract-skill.sh includes basic security checks to prevent path traversal. The instructions in SKILL.md and the OpenClaw hooks are transparently aligned with the stated goal of iterative learning and do not contain evidence of malicious prompt injection, data exfiltration, or unauthorized persistence.
能力评估
Purpose & Capability
Name/description (self-improvement / capture learnings) align with the code and documentation. Scripts, templates, and hook handlers are all about logging learnings, producing reminders, and scaffolding new skills — nothing unrelated (no cloud credentials, no system-wide config edits beyond copy-to-home instructions).
Instruction Scope
SKILL.md instructs the user to create .learnings files, optionally copy/enable an OpenClaw hook, and promote learnings into workspace files (SOUL.md, AGENTS.md, etc.). The included activator and error-detector scripts only print reminder blocks and the error-detector reads CLAUDE_TOOL_OUTPUT to detect failures. This is consistent with the purpose, but promoting arbitrary content into workspace files will change future agent context (i.e., it can act as prompt/context injection) — the user should review learnings before promoting them globally.
Install Mechanism
No automatic install spec or remote download. This is effectively an instruction-only skill with helper scripts and hook handlers included in the bundle. The extract-skill.sh has safety checks (refuses absolute/.. output dirs) and the scripts are simple and local. There are no URLs that fetch/extract arbitrary archives at install time.
Credentials
The skill declares no required env vars or credentials. The error-detector script reads CLAUDE_TOOL_OUTPUT (a platform-provided tool output variable) which is reasonable for detecting tool errors. No secrets or unrelated environment access are requested.
Persistence & Privilege
always is false and the skill is user-invocable; the hook handlers inject a virtual reminder at agent bootstrap when the hook is installed/enabled. That behavior is expected for a reminder hook but it does give recurring presence if you copy/enable the hook in ~/.openclaw/hooks — review before enabling globally. Autonomous invocation is allowed by default (normal for skills) but not excessive here.
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 wangkang-skill - 安装完成后，直接呼叫该 Skill 的名称或使用
/wangkang-skill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
v1.0.1
No changes detected in this release.
- Version updated to 1.0.3 with no modifications to files or documentation.
v1.0.2
Self-improvement skill v1.0.2 introduces structured learning and error logging for continuous improvement.
- Added comprehensive documentation on when and how to log learnings, errors, and feature requests using markdown files.
- Defined standardized log entry formats for learnings, errors, and feature requests, with clear metadata and promotion workflow.
- Provided quick reference tables for actions to take in different situations (failures, user corrections, feature requests, knowledge gaps, best practices, etc).
- Detailed OpenClaw integration steps, including workspace structure, installation, optional hooks, and promotion targets for important learnings.
- Included fallback instructions for use with other agent frameworks outside OpenClaw.
- Outlined process for promoting key learnings to project-wide memory files such as CLAUDE.md, AGENTS.md, and TOOLS.md.
元数据
常见问题
wangkang-skill 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 271 次。
如何安装 wangkang-skill？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install wangkang-skill」即可一键安装，无需额外配置。
wangkang-skill 是免费的吗？
是的，wangkang-skill 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
wangkang-skill 支持哪些平台？
wangkang-skill 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 wangkang-skill？
由 wangkang5（@wangkang5）开发并维护，当前版本 v1.0.1。
推荐 Skills