← 返回 Skills 市场
148
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install skill-kit
功能描述
Claude Code skill management. writer - create new skills [writer.md], lint - validate and fix frontmatter [lint.md], merge - combine related skills [merge.md...
安全使用建议
This skill is designed to manage other skills and therefore reads and modifies many Claude runtime files (~/.claude/skills, ~/.claude/agents, ~/.claude/settings.json, ~/.claude/hooks, ~/.claude/data). Before running it: 1) Inspect scripts/trigger-compile.sh and any triggers declared in this SKILL.md to see exactly what hook scripts would be generated. 2) Run trigger compile with the provided dry-run flag first to preview changes. 3) Backup ~/.claude/settings.json and any skill directories you care about (cp or git) so you can restore if hooks or SKILL.md are changed. 4) Expect operations that move files to ~/.claude/.bak and that may edit SKILL.md frontmatter (the upgrade/dedup flows describe auto-editing in places). 5) Treat this skill as high-privilege: do not run compile/upgrade actions without confirming each proposed change, and avoid enabling it for automatic/always-on execution unless you trust the source and have reviewed its scripts.
功能分析
Type: OpenClaw Skill
Name: skill-kit
Version: 0.1.2
The `skill-kit` bundle provides a suite of tools for managing AI agent skills, most notably the `trigger-compile.sh` script which generates executable bash hooks and modifies the global `settings.json` configuration. This mechanism creates a significant attack surface by turning markdown frontmatter into executable code. Additionally, the 'Ralph Mode' described in `SKILL.md` and the automated dependency detection in `upgrade.md` involve broad file system access and the potential to bypass user confirmation for certain actions. While these features are aligned with the stated purpose of skill management, the high-privilege operations and automated code generation are inherently risky.
能力评估
Purpose & Capability
The name/description (skill creation, linting, merging, trigger compilation) match the behavior in SKILL.md and the included scripts: scanning ~/.claude/skills, creating SKILL.md, moving skill/agent files, and compiling triggers. Requests and actions are broadly proportional to a skill-management toolkit.
Instruction Scope
Runtime instructions and the included script read and write many user-global paths (~/.claude/skills, ~/.claude/plugins, ~/.claude/agents, ~/.claude/settings.json, ~/.claude/hooks, ~/.claude/data) and perform moves (mv), backups, and edits to SKILL.md and settings.json. The workflow includes automatic 'depends-on' edits, backups, and moving agent files — these are expected for a manager tool but are high-impact operations and some steps are described as 'required' (auto-modify frontmatter) while other places mandate user confirmation, creating ambiguity about what will run without explicit approval.
Install Mechanism
No installer — instruction-only with one included shell script (scripts/trigger-compile.sh). There is no remote download, which lowers supply-chain risk; however the shipped shell script writes auto-generated hook scripts to ~/.claude/hooks and overwrites settings.json entries via jq, so running the script performs filesystem and config mutations.
Credentials
The skill requests no environment variables or external credentials (good). It nonetheless requires broad filesystem access to user-specific Claude runtime directories — appropriate for a skill manager but sensitive because it touches user agents, plugin areas, and global settings.
Persistence & Privilege
The trigger compilation flow writes persistent hook scripts (~/.claude/hooks/trigger-*.sh), sets flag files (~/.claude/data/trigger-stop-*) and updates ~/.claude/settings.json to register hooks that can run on PreToolUse/PostToolUse/Stop. Registered hooks can 'suggest', 'block' (exit 1) or 'inject' systemMessage JSON, giving lasting control over agent execution flow. While this is aligned with the 'trigger' feature, it is a powerful privilege and could be abused if a skill registers malicious triggers or auto-modifies other skills without clear user consent.
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 skill-kit - 安装完成后，直接呼叫该 Skill 的名称或使用
/skill-kit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
v0.1.2
Enhance upgrade and route topics, add version bump procedure documentation
v0.1.1
Add route topic for topic placement recommendations, improve upgrade procedure
v0.1.0
Initial release: renamed from skill-toolkit for shorter slug
元数据
常见问题
Skill Kit 是什么？
Claude Code skill management. writer - create new skills [writer.md], lint - validate and fix frontmatter [lint.md], merge - combine related skills [merge.md... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 148 次。
如何安装 Skill Kit？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install skill-kit」即可一键安装，无需额外配置。
Skill Kit 是免费的吗？
是的，Skill Kit 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Skill Kit 支持哪些平台？
Skill Kit 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 Skill Kit？
由 es6kr（@drumrobot）开发并维护，当前版本 v0.1.2。
推荐 Skills