菜单
Install a skill into Claude Code. Use when the user wants to add a new skill, install a skill from a local folder, or set up a custom skill. Triggers on "install skill", "add skill", "set up skill", "load skill".
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | install-skill |
| description | Install a skill into Claude Code. Use when the user wants to add a new skill, install a skill from a local folder, or set up a custom skill. Triggers on "install skill", "add skill", "set up skill", "load skill". |
You are helping the user install a skill into Claude Code. Follow these steps precisely.
Ask the user where the skill is. It can be:
If it's a local folder, read the contents:
SKILL.md — this is the entry pointscripts/, templates/, or other supporting filesSKILL.md fully to understand what the skill does and what it needsIf it's a git repo, clone it to a temp directory first:
git clone <url> "$TMPDIR/skill-to-install"
From SKILL.md, identify:
node, python, curl, etc.)Verify each:
# Check env vars
echo $ENV_VAR_NAME
# Check CLI tools
which node
which python3
# Check packages
node -e "require('package-name')" 2>/dev/null && echo "OK" || echo "MISSING"
If anything is missing, tell the user what's needed and how to install it. Do NOT proceed until dependencies are satisfied.
Copy the skill folder to the personal skills directory.
⚠️ 关键规则 — 避免嵌套目录：
cp -r 源目录/ 目标目录 如果目标目录已存在，cp 会把源目录复制到目标目录里面，产生 skill-name/skill-name/ 嵌套！find 验证文件结构，不能只靠回忆复述文件名。macOS / Linux / Windows (bash):
# 先清理旧版本（避免嵌套）
rm -rf "$HOME/.claude/skills/<skill-name>"
# 再复制
cp -r /path/to/skill-folder "$HOME/.claude/skills/<skill-name>"
Windows (WSL):
rm -rf "$HOME/.claude/skills/<skill-name>"
cp -r /path/to/skill-folder "$HOME/.claude/skills/<skill-name>"
Windows (PowerShell):
if (Test-Path "$env:USERPROFILE\.claude\skills\<skill-name>") { Remove-Item -Recurse -Force "$env:USERPROFILE\.claude\skills\<skill-name>" }
Copy-Item -Recurse "C:\path\to\skill-folder" "$env:USERPROFILE\.claude\skills\<skill-name>"
如果源目录名与 <skill-name> 不一致，需要重命名：
rm -rf "$HOME/.claude/skills/<skill-name>"
cp -r /path/to/源目录名 "$HOME/.claude/skills/<skill-name>"
If the skill requires environment variables that aren't set:
macOS / Linux (zsh):
echo 'export VAR_NAME="value"' >> ~/.zshrc
source ~/.zshrc
macOS / Linux (bash):
echo 'export VAR_NAME="value"' >> ~/.bashrc
source ~/.bashrc
Windows (PowerShell):
[Environment]::SetEnvironmentVariable("VAR_NAME", "value", "User")
Tell the user they may need to restart their terminal or Claude Code for env vars to take effect.
find 列出实际文件结构（不要只用 ls，不要自己回忆文件名）：find "$HOME/.claude/skills/<skill-name>" -type f
<skill-name>/<skill-name>/ 结构，说明安装出错了，需要重新执行 Step 4：# 如果检测到嵌套，修复：
rm -rf "$HOME/.claude/skills/<skill-name>"
cp -r /path/to/skill-folder/. "$HOME/.claude/skills/<skill-name>"
ls -la "$HOME/.claude/skills/<skill-name>/scripts/"
# 如果 .sh 文件不可执行（macOS/Linux）：
chmod +x "$HOME/.claude/skills/<skill-name>/scripts/"*.sh
# 根据实际脚本文件名调整，查看 SKILL.md 中的 Usage 部分
node "$HOME/.claude/skills/<skill-name>/scripts/search.mjs" --help
# 或
python "$HOME/.claude/skills/<skill-name>/scripts/main.py" --help
Restart Claude Code (exit and re-run claude) for the new skill to be recognized.
Tell the user to try invoking it with /<skill-name> or by asking a question that matches the skill's description.
After installation, report to the user:
✅ Skill "<skill-name>" installed successfully!
📁 Location: $HOME/.claude/skills/<skill-name>/
🔧 Dependencies: <list what was checked/set up>
🔑 Env vars: <list what was configured>
💬 Invoke with: /<skill-name> or ask about <description>
⚠️ Restart Claude Code to pick up the new skill.
LLM 微调专家。使用场景：微调 LLMs、训练自定义模型、适配基础模型进行特定任务。用于配置 LoRA/QLoRA 适配器、准备 JSONL 训练数据集、设置超参数、适配器训练、迁移学习、Hugging Face PEFT 微调、指令微调、RLHF、DPO 或量化部署微调模型。触发词：LoRA、QLoRA、PEFT、微调、适配器调优、LLM 训练、模型训练、自定义模型。
深度研究工具。仅在用户明确说'深度研究'、'深入调查'、'全面报告'时使用。比普通搜索慢 10-100 倍。对于普通研究/查找请求，使用 parallel-web-search。支持多轮：传递 --previous-interaction-id 以继续之前的上下文。
AI 模型提示工程指南：LLM、图像生成器、视频模型。技术：思维链、少样本、系统提示、负提示。模型：Claude、GPT-4、Gemini、FLUX、Veo、Stable Diffusion 提示。用于：更好的 AI 输出、一致的结果、复杂任务、优化。触发词：提示工程、如何提示、更好的提示、提示技巧、提示指南、LLM 提示、图像提示、AI 提示、提示优化、提示模板、提示结构、有效提示、提示技术。
RAG（检索增强生成）模式专家。用于构建 RAG 管道、嵌入文档、实现混合搜索、上下文检索、HyDE、Agentic RAG、多模态 RAG、查询分解、重排或 pgvector 搜索。触发词：RAG、检索、LLM、上下文、嵌入、HyDE、重排、pgvector、多模态。
Apache Spark 作业优化：分区策略、缓存、shuffle 优化和内存调优。用于提升 Spark 性能、调试慢作业或扩展数据处理管道。触发词：Spark、性能优化、分区、缓存、shuffle、数据倾斜。
Web 研究工具：搜索多个来源、综合发现并生成带引用的研究报告。用于用户要求在线研究主题、搜索网页、查找信息、获取最新信息、对比选项或生成研究报告。触发词：研究、网上搜索、查找信息、最新信息、对比选项、研究报告。