Due Diligence Pipeline Skill

You are orchestrating a comprehensive due diligence research workflow. Follow these phases exactly.

Phase 0: Gather User Inputs

FIRST, use AskUserQuestion to collect ALL required information before starting any research:

Use AskUserQuestion with these questions:

1. Question: "What is the company name you want to research?"
   Header: "Company"
   Options:
   - Label: "Enter company name" / Description: "Type the company name in 'Other'"

2. Question: "Which languages should the final deliverables be translated into?"
   Header: "Languages"
   multiSelect: true
   Options:
   - Label: "Traditional Chinese (Taiwan)" / Description: "繁體中文 - Taiwan market"
   - Label: "Simplified Chinese" / Description: "简体中文 - Mainland China market"
   - Label: "Japanese" / Description: "日本語 - Japan market"
   - Label: "Korean" / Description: "한국어 - Korea market"

3. Question: "What is the investment thesis or focus area?"
   Header: "Focus"
   Options:
   - Label: "General overview" / Description: "Comprehensive company analysis"
   - Label: "Market opportunity" / Description: "TAM/SAM/SOM and competitive landscape"
   - Label: "Technology moat" / Description: "Technical differentiation and IP"
   - Label: "Team & execution" / Description: "Leadership and operational capability"

4. Question: "Do you have local files in data-room/ to include?"
   Header: "Data Room"
   Options:
   - Label: "Yes, analyze them" / Description: "Include files from data-room/ folder"
   - Label: "No, web research only" / Description: "Skip local file analysis"

Store the responses and proceed to Phase 1.

Phase 1: Multi-Agent Research (Parallel)

Create tasks and spawn multiple research agents IN PARALLEL using the Task tool:

Research Agents to Spawn (all parallel):

1. •

   Market Research Agent

   • •subagent_type: voltagent-research:market-researcher
   • •Prompt: Research market size, growth trends, competitive landscape for [COMPANY]
2. •

   Competitive Analysis Agent

   • •subagent_type: voltagent-research:competitive-analyst
   • •Prompt: Analyze [COMPANY]'s competitors, market positioning, SWOT analysis
3. •

   Trend Analysis Agent

   • •subagent_type: voltagent-research:trend-analyst
   • •Prompt: Identify industry trends, emerging opportunities, risks for [COMPANY]'s sector
4. •

   Data Research Agent

   • •subagent_type: voltagent-research:data-researcher
   • •Prompt: Gather financial data, funding history, key metrics for [COMPANY]

If user selected "Yes" for data-room files, also spawn:

1. •Local File Analysis Agent
   • •subagent_type: Explore
   • •Prompt: Analyze all files in data-room/ directory, extract key insights about [COMPANY]

IMPORTANT: Launch all research agents in a SINGLE message with multiple Task tool calls.

Phase 2: Compile Draft Report

After all research agents complete, compile findings into a structured report:

Write to: reports/[COMPANY]-due-diligence-draft.md

Report structure:

# [COMPANY] Due Diligence Report

## Executive Summary
[Synthesize key findings]

## Company Overview
[From research agents]

## Market Analysis
[TAM/SAM/SOM, market trends]

## Competitive Landscape
[Key competitors, positioning]

## Technology & Product
[Technical moat, differentiation]

## Team & Leadership
[Key personnel, track record]

## Financial Overview
[Funding, metrics, projections]

## Risk Assessment
[Key risks and mitigations]

## Investment Thesis
[Based on user's focus area]

## Appendix
[Data sources, methodology]

Phase 3: Fact-Check (Multi-Agent)

Spawn fact-check agents to validate the report:

1. •

   Fact-Check Agent

   • •subagent_type: fact-check-agent
   • •Prompt: Review reports/[COMPANY]-due-diligence-draft.md against source documents in data-room/. Validate claims, identify logical fallacies, flag unsupported assertions.
2. •

   Research Analyst Validation

   • •subagent_type: voltagent-research:research-analyst
   • •Prompt: Cross-verify key claims in reports/[COMPANY]-due-diligence-draft.md using web research. Flag any discrepancies.

Launch both in parallel. After completion, update the report with corrections.

Write final report to: reports/[COMPANY]-due-diligence-final.md

Phase 4: Generate VC Blog Post

Create a VC-style blog post based on the report:

Write to: reports/[COMPANY]-vc-blog.md

Blog structure:

# [Compelling Headline About Company/Market]

*By [Your Fund Name] | [Date]*

## The Opportunity

[Hook - why this matters]

## What [COMPANY] Does

[Clear, accessible explanation]

## Why Now

[Market timing, trends]

## Why [COMPANY]

[Competitive advantages]

## Our Take

[Investment thesis in accessible language]

---
*Disclaimer: This is for informational purposes only and does not constitute investment advice.*

Phase 5: Translate Report (Multi-Agent Parallel)

For EACH language selected by user, spawn a translation-agent:

Task tool with:
- subagent_type: translation-agent
- Prompt:
  Source: English
  Target: [Selected Language]
  Region: [Appropriate region]
  Style: formal/professional

  [Contents of reports/[COMPANY]-due-diligence-final.md]

Output files:

• •reports/[COMPANY]-due-diligence-zh-TW.md (Traditional Chinese)
• •reports/[COMPANY]-due-diligence-zh-CN.md (Simplified Chinese)
• •reports/[COMPANY]-due-diligence-ja.md (Japanese)
• •reports/[COMPANY]-due-diligence-ko.md (Korean)

Launch ALL translation agents in parallel in a single message.

Phase 6: Translate Blog (Multi-Agent Parallel)

Same as Phase 5, but for the blog post:

Output files:

• •reports/[COMPANY]-vc-blog-zh-TW.md
• •reports/[COMPANY]-vc-blog-zh-CN.md
• •reports/[COMPANY]-vc-blog-ja.md
• •reports/[COMPANY]-vc-blog-ko.md

Launch ALL blog translation agents in parallel in a single message.

Phase 7: Summary Report

After all phases complete, provide user with:

# Pipeline Complete: [COMPANY] Due Diligence

## Deliverables Created

### Reports
- [ ] `reports/[COMPANY]-due-diligence-final.md` (English)
- [ ] `reports/[COMPANY]-due-diligence-zh-TW.md` (繁體中文)
- [ ] ... (list all created files)

### Blog Posts
- [ ] `reports/[COMPANY]-vc-blog.md` (English)
- [ ] ... (list all translations)

## Research Summary
[Brief summary of key findings]

## Fact-Check Results
[Any corrections made]

## Next Steps
- Review deliverables
- [Suggested follow-up actions]

Execution Notes

1. •Always maximize parallelism: Launch independent agents together
2. •Track progress: Use TaskCreate/TaskUpdate to show progress
3. •Handle failures gracefully: If an agent fails, note it and continue
4. •Respect user inputs: Use exactly what user specified
5. •Write files as you go: Don't wait until the end

Quick Start

When user runs /ddp:

1. •Immediately start Phase 0 (AskUserQuestion)
2. •Proceed through phases sequentially, but parallelize within phases
3. •Provide progress updates between phases