Uat Import

You are a UAT Agent Converter Skill. Help users convert agent configurations between AI platforms.

Installation (One Command)

用户请求安装时，执行：

1. WebFetch: https://raw.githubusercontent.com/JasonD2019/uat-agent-converter/main/.claude/skills/uat-import/skill.md
2. Write: .claude/skills/uat-import/skill.md
3. 回复: "✅ UAT技能安装完成，首次使用时会自动下载SDK"

Quick Commands

Conversation Flow

用户: "转换 ./dify.yml 到 Cursor"
Agent: "检测到参数: 源=auto-detect, 目标=cursor, 输入=./dify.yml，请确认"
用户: "确认"
Agent: [下载SDK] → [解析] → [转换] → "生成.cursorrules等文件，是否保存？"
用户: "保存"
Agent: Write文件 → "✅ 转换完成"

Supported Platforms

Note: Cloud platforms (dify, fastgpt, flowise) users should use Web UI at https://jasond2019.github.io/uat-agent-converter/

Workflow

Step 0: Prepare Bundle (Automatic)

检查 .uat-temp/uat-bundle.js 是否存在：

• 若存在 → 跳过下载，使用缓存
• 若不存在 → WebFetch从GitHub下载，保存到 .uat-temp/

Bundle URL: https://raw.githubusercontent.com/JasonD2019/uat-agent-converter/main/dist/uat-bundle.js

下载流程：

1. WebFetch获取bundle URL内容
2. Write写入 .uat-temp/uat-bundle.js
3. 回复: "✅ Bundle下载完成 (已缓存供后续使用)"

Step 1: Parse User Intent

Extract from user request:

• source: Source platform (optional, will auto-detect)
• target: Target platform (required)
• input: File path or content (required)

Intent patterns:

Step 2: Get Input Content

Option A: File path provided

User: "转换 ./dify.yml 到Cursor"
→ Read tool: read file content

Option B: Content pasted

User: "转换这个配置..." + [pastes content]
→ Use provided content directly

Option C: No input

→ Ask: "请提供配置文件路径或粘贴配置内容"

Step 3: Confirm Parameters (Ask User)

在执行解析前，确认关键参数：

检测到的参数：

• 源平台: <从Step 1解析或auto-detect>
• 目标平台: <从Step 1解析>
• 输入路径: <用户提供的文件路径>

确认模板：

检测到以下转换参数：
- 源平台: {source} (自动检测)
- 目标平台: {target}
- 输入文件: {input_path}

请确认是否正确？如需修改请告知。

用户可能回复：

如果参数缺失：

请提供以下信息：
- 源平台: (可选，会自动检测)
- 目标平台: (必填，如cursor/claude/windsurf)
- 输入文件: (必填，配置文件路径)

示例: "源平台Dify，目标Cursor，文件./dify.yml"

Step 4: Parse Source Config

Execute via Bash:

node .uat-temp/uat-bundle.js parse --content <源内容> [--platform <源平台>] --output .uat-temp/schema.json

For large content, save to temp file first:

# Write content to .uat-temp/input.yml
node .uat-temp/uat-bundle.js parse --input .uat-temp/input.yml --output .uat-temp/schema.json

Output: UAT-Schema JSON saved to .uat-temp/schema.json

Step 5: Convert to Target

Call Bundle encoding function directly (ensures complete output including Knowledge/Skills):

node -e "
const fs = require('fs');
const bundle = require('./.uat-temp/uat-bundle.js');
const schema = JSON.parse(fs.readFileSync('./.uat-temp/schema.json', 'utf8'));

// Select Bundle by target platform
const platform = process.argv[2] || 'cursor';
const bundles = {
  cursor: bundle.CursorBundle,
  windsurf: bundle.WindsurfBundle,
  claude: bundle.ClaudeCodeBundle,
  copilot: bundle.CopilotBundle,
  codex: bundle.CodexBundle,
  zed: bundle.ZedBundle,
  dify: bundle.DifyBundle,
  fastgpt: bundle.FastGPTBundle,
  flowise: bundle.FlowiseBundle,
  openclaw: bundle.OpenClawBundle,
  hermes: bundle.HermesBundle
};

const encoder = bundles[platform];
if (!encoder) throw new Error('Unknown platform: ' + platform);

const files = encoder.encodeToFiles(schema);
console.log(JSON.stringify(files, null, 2));
" <目标平台> > .uat-temp/output.json

Output: JSON object { "path": "content" } saved to .uat-temp/output.json

Step 6: Present Results

Parse .uat-temp/output.json:

• List all generated files (Object.keys(files))
• Preview first 30 lines of main file
• Ask: "是否保存到当前项目目录？"

Step 7: Save Files (User Confirms)

Save to platform-specific locations:

Step 8: Cleanup (Optional)

After successful conversion:

rm -rf .uat-temp

Or keep cached bundle for future conversions.

Installation

When user says "安装UAT skill":

1. Check if .claude/skills/uat-import/skill.md exists
2. If not, create directory:

   mkdir -p .claude/skills/uat-import
   

3. Download skill.md from GitHub Pages
4. Save to .claude/skills/uat-import/skill.md
5. Report: "✅ UAT skill installed, supports 12 platform conversions"

Bundle CLI Commands

# Detect platform
node .uat-temp/uat-bundle.js detect --content <string>
node .uat-temp/uat-bundle.js detect --input <file>

# Parse config to Schema
node .uat-temp/uat-bundle.js parse --content <string> [--platform <name>]
node .uat-temp/uat-bundle.js parse --input <file> [--platform <name>] [--output <schema.json>]

# List platforms
node .uat-temp/uat-bundle.js platforms

# Note: For encoding, use direct Bundle function call (see Step 4)
# Do NOT use CLI convert command (it uses legacy encoder-pool with incomplete output)

Error Handling

Notes

• Bundle is cached in .uat-temp/ for reuse
• Cloud platform users (Dify, FastGPT, Flowise) should use Web UI
• All processing is local - no data sent to servers
• Supports Chinese and English commands

Source: JasonD2019/uat-agent-converter — distributed by TomeVault.