21pounder/terminalagent/deep-research

Deep Research (Dify Powered)

This skill delegates research tasks to a specialized Dify Workflow that:

1. Searches official documentation, tutorials, and academic resources
2. Analyzes the topic with DeepSeek Reasoner for deep insights
3. Iteratively searches for examples, solutions, and related research
4. Generates a comprehensive research report with structured findings

Use this skill for:

• Code implementation research (APIs, libraries, best practices)
• Academic paper analysis and literature review
• Technology comparison and evaluation
• Any topic requiring comprehensive, structured research

Parameters

{
  "type": "object",
  "properties": {
    "coding_task": {
      "type": "string",
      "description": "The coding task or question to research (required)"
    },
    "tech_stack": {
      "type": "string",
      "description": "Technology stack context (e.g., 'React 18, TypeScript, Next.js')"
    },
    "depth": {
      "type": "integer",
      "minimum": 1,
      "maximum": 5,
      "default": 3,
      "description": "Research depth (1=quick, 3=standard, 5=comprehensive)"
    }
  },
  "required": ["coding_task"]
}

Workflow

Step 1: Locate the Client Script

The client script is at: .claude/skills/deep-research/scripts/dify-client.ts Use Glob to find the absolute path if needed.

Step 2: Execute Research

Use Bash to run the script with npx tsx:

npx tsx "<path_to_script>" "<coding_task>" "<tech_stack>" <depth>

Examples:

# Basic research
npx tsx ".claude/skills/deep-research/scripts/dify-client.ts" "How to implement OAuth2 in Next.js"
 
# With tech stack
npx tsx ".claude/skills/deep-research/scripts/dify-client.ts" "Add real-time notifications" "React 18, Socket.io"
 
# With depth
npx tsx ".claude/skills/deep-research/scripts/dify-client.ts" "Optimize database queries" "PostgreSQL, Prisma" 5

Step 3: Process Results

The script returns JSON with two main fields:

• analysis: Task analysis with knowledge gaps and tech requirements
• guide: Complete implementation guide with:
• TASK_SUMMARY
• DEPENDENCIES (install commands)
• FILES_TO_CREATE (complete code)
• FILES_TO_MODIFY (change instructions)
• ENVIRONMENT_CONFIG
• VERIFICATION (test commands)
• GOTCHAS (common issues)
• SOURCES

Step 4: Implement

Use the guide to:

1. Install dependencies
2. Create new files with provided code
3. Modify existing files as instructed
4. Set up environment variables
5. Run verification commands

Error Handling

If the script returns an error:

1. The script automatically loads .env from the project root (no manual env setup needed)
2. If still failing, check that .env contains DIFY_API_KEY and DIFY_BASE_URL
3. Verify the Dify API is accessible

Output Format

{
  "success": true,
  "data": {
    "analysis": "{ JSON task analysis }",
    "guide": "# Implementation Guide\n\n### TASK_SUMMARY\n..."
  },
  "metadata": {
    "duration_ms": 45000,
    "workflow_run_id": "abc123"
  }
}

When to Use

• Complex implementation tasks requiring research
• Integrating unfamiliar APIs or libraries
• Finding best practices and working examples
• Understanding new frameworks or patterns