Vc Diligence

VC Diligence Skill: The Ultimate Lead Partner Framework (V26 - Master Synthesis & Serper Edition)

This is the definitive handbook for executing the vc-diligence skill. It combines Recursive Hypothesis-Driven Discovery with Hyper-Depth Content Harvesting, Specialized Architecture Auditing, Product-Market Fit (PMF) Leveling, and Excalidraw Visual Mastery. The goal is to produce an intimate, 30+ page master audit that identifies the technical plumbing, commercial moats, and legal red-flags of any startup.

0. PREREQUISITES & SETUP (CRITICAL)

Before executing this skill, ensure the following dependencies are installed to support diagram rendering:

1. Python Playwright: pip install playwright
2. Browser Binaries: npx playwright install chromium
3. Internal Renderer: This skill depends on the excalidraw-diagram skill's renderer located at /Users/harrisonluo/.agents/skills/excalidraw-diagram/references/render_excalidraw.py.
4. Specialized Agents:
   • Use architecture-auditor for Phase 4 & 10.5 (Technical Teardown).
   • Use excalidraw-diagram for Phase 12 (Diagram Generation).
5. API Keys: Ensure SERPER_API_KEY is available in the environment for high-fidelity image discovery.

1. THE "HYPER-DEPTH" ANALYST MINDSET

1.1 The 30-Page Master Mandate

A brief summary is a failure. You are building an exhaustive legal and commercial case.

• The Isomorphism Principle: Diagrams must convey business logic through structure, not just labels. If you remove the text, the shape should still be the meaning.
• The Education Test: Could someone learn something concrete from the diagram? It must show actual formats, real event names, and concrete examples.
• The Traceability Rule: Every claim MUST be backed by a link or raw finding file.
• Fact-Dense Writing: Omit fluff. Use metrics, specific library names, and exact pricing.
• Exhaustive Extraction: Use web_fetch on 10-20 sources per wave. Snippets are insufficient. You MUST extract full text to construct detailed feature breakdowns.
• The Search-to-Fetch Ratio: For every Turn that includes a google_web_search, you MUST execute a corresponding web_fetch Turn for the top 3-5 high-signal results.

1.2 The "Proxy Research" Principle

If direct information about a startup is scarce (e.g., stealth), you MUST research the ecosystem proxies:

• Technology Proxies: Research the benchmarks of the specific tools/frameworks they use.
• Personnel Proxies: Research the "Success Patterns" of the companies the founders came from.
• Competitor Proxies: Perform deep teardowns of incumbents to identify the exact "White Space" the startup is filling.

2. TABLE OF CONTENTS

1. Phase 0: Workspace Architecture & Initialization
2. Phase 1: Objective-Oriented Recursive Search & Content Harvesting
3. Phase 2: MCP Service Discovery & Third-Party Intelligence
4. Phase 3: The Founder Alpha & Team Audit
5. Phase 4: Specialized Technical Teardown (Architecture Auditor)
6. Phase 5: Product-Market Fit (PMF) Audit & Leveling
7. Phase 6: Commercial, Financial & Unit Economic Diligence
8. Phase 7: Market Dynamics, Competitive Moats & Exit Benchmarks
9. Phase 8: The 2025-2026 AI Impact & Industry Trend Analysis
10. Phase 9: The Master VC Diligence Questionnaire (Exhaustive)
11. Phase 10: Legal, Regulatory & Governance Risk Auditing
12. Phase 11: Visual Asset Acquisition (Serper Image Search)
13. Phase 11.5: Formal Visual Design Specification
14. Phase 12: Synthesis - Master Deliverables Protocol
15. Phase 12.5: The Iterative Master Expansion Step
16. Phase 12.6: Technical Context Gap-Filling (Deep Dive Expansion)
17. Phase 13: Advanced Visual Argumentation (Skill Delegation)
18. Phase 13.1: Visual Quality Assurance & Refinement Loop
19. Phase 14: Formatting Audit (Tables vs. Bullets)
20. Phase 15: Final Holistic Synthesis & Conviction Scoring
21. Phase 16: Final Quality Assurance & Hand-off

3. PHASE 0: WORKSPACE ARCHITECTURE & INITIALIZATION

mkdir -p diligence_[startup_name]/raw_findings/team
mkdir -p diligence_[startup_name]/raw_findings/tech
mkdir -p diligence_[startup_name]/raw_findings/market
mkdir -p diligence_[startup_name]/raw_findings/legal
mkdir -p diligence_[startup_name]/diagrams
mkdir -p diligence_[startup_name]/deliverables
mkdir -p diligence_[startup_name]/assets

4. PHASE 1: OBJECTIVE-ORIENTED RECURSIVE SEARCH & CONTENT HARVESTING

4.1 The Wave-Based Discovery Objectives (Hyper-Depth)

Wave 0: Product DNA & User Persona (The Foundation)

• Central Question: "What exactly is the product, how does it function for the end-user, and who are the core personas it serves?"
• Guiding Mandate: Understand the 'Jobs to be Done' before auditing the 'Right to Win'.
• Discovery Specifics: Feature inventory, user personas (Economic Buyer vs. Daily User), UX/Workflow mapping, and core value prop.
• Example Evidence: A 'How it Works' page; a user manual; onboarding tutorials; detailed customer testimonials.

Wave 0.5: Product-Market Fit (PMF) Audit

• Central Question: "At what level of PMF is this startup currently operating (Level 1–4), and what are the signals of satisfaction, demand, and efficiency?"
• The PMF Method (First Round Framework):
  • Level 1 (Nascent): Solving a problem for 3-5 customers. Qualitative: "The Grind."
  • Level 2 (Developing): Scalable sales channel emerging. Qualitative: "Lite PMF."
  • Level 3 (Strong): Floodgates open. Efficiency + GTM refinement. Qualitative: "The Geyser."
  • Level 4 (Extreme): Categorical dominance. Global expansion.
• Discovery Specifics: Retention/Usage depth (Satisfaction), Sales velocity/Inbound (Demand), Magic Number/Burn Multiple (Efficiency).

Wave 1: Identity, Trajectory & Capital Structure

• Central Question: "Who is this company at its core, how much capital have they consumed, and what is their growth velocity?"
• Discovery Specifics:
  • Legal Structure: Identify parent entity, DBAs, and former names.
  • Cap Table Proxies: Hunt for funding announcements and SEC Form D filings.
  • Hiring Velocity: Search for "LinkedIn Headcount Growth" patterns or historical job postings.
• Example Evidence: A Press Release confirming funding rounds; a Crunchbase entry showing round intervals.

Wave 2: Founder Alpha & Talent Density

• Central Question: "Why is this specific team uniquely qualified to win this category, and what are their individual 'superpowers'?"
• Discovery Specifics:
  • The 'Spike': Identify the world-class expertise of each founder (Academic, Serial, BigTech).
  • Social Graph: Search for testimonials, board seats, or past managers.
  • Conflict/Cohesion: Find evidence of past collaboration (worked together at X for Y years).
• Example Evidence: Google Scholar profiles; LinkedIn recommendations; academic publication history.

Wave 3: Technical Architecture & Product Moat

• Central Question: "Exactly how is this built, where are the structural bottlenecks, and is it a defensible moat or a thin wrapper?"
• MANDATORY: Use web_fetch on all technical docs, API references, and engineering blogs found. Build a component-by-component feature teardown.
• Discovery Specifics:
  • Infrastructure Stack: Identify the 'Plumbing' (AWS/GCP, DBs, LLM providers).
  • Proprietary IP: Search for internal project names, custom protocols, or patents.
  • Performance Benchmarks: Hunt for latency, throughput, or accuracy metrics.
• Example Evidence: API references; GitHub dependency audits; technical whitepapers.

Wave 4: Market Dynamics & Competitive Displacement

• Central Question: "Which $B+ market is being eaten, who are the incumbents being displaced, and what are the exit benchmarks?"
• The Synthesis Mandate: Do NOT search for "TAM SAM SOM." Search for Atomic Facts:
  • Market Size Proxies: "Total number of [Target User Profile]" or "Total addressable market of [Industry Sector]".
  • Pricing Proxies: "[Competitor] pricing" or "[Incumbent] revenue per user".
  • Adoption Proxies: "[Competitor] customer count".
• Hypothesis-First (MANDATORY): Based on Waves 1-3, define the disrupted market and name the likely incumbents and emerging challengers.
• Discovery Specifics:
  • Incumbent Vulnerability: Search for "Negative Reviews" or "Integrations Issues" with incumbents.
  • Category Pricing: Hunt for competitor pricing pages or 'Talk to Sales' leakages.
  • Exit Comps: Find 3-5 recent M&A deals in the sector.
• Example Evidence: A Forrester Wave report showing the incumbent's market share is declining; an M&A deal at a 12x revenue multiple.

4.2 Query Fanout Mandate

For every Central Question or Hypothesis, you MUST generate a fanout of 3-5 distinct, multi-angle queries.

• Angle 1: Direct Discovery: (e.g., \"[Startup] [Topic]\")
• Angle 2: Proxy Signals: Search for indirect evidence (e.g., \"[Topic] reviews\", "hiring for [Topic]").
• Angle 3: Technical/Source-Specific: Target deep sources (site:github.com, filetype:pdf).
• Angle 4: Competitive/Market Context: (e.g., \"[Startup] vs [Competitor]\").

4.3 The Search Critic & Intent SOP (Least-to-Most Protocol)

Before executing ANY query, apply Least-to-Most Prompting:

1. Decomposition: "Break the Central Question down into the fundamental sub-questions that need to be answered first."
2. Sequential Solving: "Solve each sub-question one by one, using the answer from the previous step to inform the next search angle."
3. The 'Reasonable Searchability' Check: Search for the atomic ingredients instead of the finished meal.
4. The Search-to-Fetch Mandate: For each query fanout, identify the top high-signal URLs and execute web_fetch.

5. PHASE 4: SPECIALIZED TECHNICAL TEARDOWN (ARCHITECTURE AUDITOR)

CRITICAL DELEGATION: For this phase, you MUST activate the architecture-auditor skill.

1. Harvest Ingredients: Collect real API names, JSON schemas, and method calls to prove the depth of integration.
2. Audit Moat: Use the architecture-auditor specialist to perform the "Wrapper vs. Moat" test.
3. Feature Teardown: Draft the detailed component-by-component analysis for the final report.

6. PHASE 5: PRODUCT-MARKET FIT (PMF) AUDIT & LEVELING

You MUST apply the First Round Capital PMF Method to determine the company's trajectory.

6.1 The 3 Dimensions of PMF

1. Satisfaction: Audit customer reviews, developer forums (Discord/Reddit), and case studies. Are they "Friend-zoned" (like but don't need) or is it a "Painkiller"?
2. Demand: Search for inbound interest signals, waiting lists, and pricing power (ability to raise prices without churn).
3. Efficiency: Calculate/Model the Magic Number, Burn Multiple, and CAC Payback based on revenue vs. headcount growth.

6.2 The 4 Levers (The 4Ps) Evaluation

If the company is stuck between levels, identify which lever they are pulling:

• Persona: Moving upmarket or changing the ICP.
• Problem: Pivoting to a more urgent/important problem.
• Promise: Refining the unique value prop.
• Product: Feature iteration to meet the promise.

7. PHASE 10.5: FORMAL VISUAL DESIGN SPECIFICATION (FACT-DENSE)

Before creating Excalidraw diagrams, you MUST write a Visual Design Specification that formalizes the argument. Use the architecture-auditor skill to ensure the spec is technical and accurate.

7.1 The Research Mandate (Visual Ingredients)

For each diagram, you must list the Evidence Artifacts discovered in research:

• Architecture: List the actual API endpoints, event names, and model names.
• Market: List the specific market share percentages and incumbent names.
• Economic: List the exact ARR milestones and gross margin percentages.

8. PHASE 11: VISUAL ASSET ACQUISITION (SERPER IMAGE SEARCH)

If google_web_search fails to provide direct image URLs, you MUST use the Serper API for high-fidelity image discovery.

8.1 Serper Image Discovery SOP

1. Search Protocol: Use run_shell_command to call the Serper Images API.
2. Command Template:

curl -X POST 'https://google.serper.dev/images' \
  -H "X-API-KEY: $SERPER_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"q": "[Founder Name] [Company] headshot", "num": 10}'

3. URL Extraction: Identify the imageUrl from the JSON response.
4. Download & Verification: Follow the standard curl -Lfg and file check protocol.

9. PHASE 12: SYNTHESIS - THE MASTER DELIVERABLES PROTOCOL

9.1 full_diligence_report.md (The 30+ Page Master Audit)

• Header 1: Executive Summary (3-paragraph distillation).
• Header 2: Company Profile & Core Product (Legal names, pivots, exhaustive feature list with sub-bullets).
• Header 3: Product-Market Fit Audit (Satisfaction/Demand/Efficiency dimensions, PMF Level 1-4 verdict).
• Header 4: The Founder Alpha Audit (Complete histories, social graphs, citations, and links).
• Header 5: Financing & Capital Structure (Round-by-round partners, valuations, board composition).
• Header 6: Detailed Technical Architecture (Plumbing, stack choices, bottlenecks, "Wrapper vs. Moat" test).
• Header 7: Market & Competitive Landscape (TAM/SAM/SOM synthesis, incumbent weaknesses, displacement economics table).
• Header 8: Commercials & Unit Economics (GTM strategy, LTV/CAC modeling, churn analysis).
• Header 9: Legal, Regulatory & Governance (IP provenance, compliance status).
• Header 10: Master Research Appendix (Categorized links and full source citations).
• Header 11: Final Investment Verdict (Detailed rationale explaining the "Why").

10. PHASE 12.5: THE ITERATIVE MASTER EXPANSION STEP (MANDATORY)

Once the initial draft of the full_diligence_report.md is complete, you MUST execute this iterative expansion cycle:

1. Section Elaborator: Add at least 3-5 verbose paragraphs per major section.
2. Structural Depth: Convert flat lists into hierarchical structures with H3/H4 sub-headings and nested sub-bullets.
3. Citation Check: Ensure every factual claim has a corresponding source link (3-5 links per section).

11. PHASE 12.6: TECHNICAL CONTEXT GAP-FILLING (DEEP DIVE EXPANSION)

After generating the initial draft of the technical_commercial_deep_dive.md, you MUST perform a Technical Gap Audit:

1. Identify "Thin" Sections: Scan the document for generic technical descriptions (e.g., "They use cloud infrastructure").
2. Reissue Deep Searches: Execute 3-5 high-resolution queries specifically targeting missing primitives identified in research:
   • Specific Algorithms: "Does [Startup] use [Specific Algorithm X] or [Variant Y]?"
   • Infrastructure Quirks: "Is [Startup] on AWS Bedrock or using raw EC2 H100s?"
   • Data Flow Detail: "What is the specific JSON payload for the [X] endpoint?"
3. Future Diligence Mapping: Add a "Technical Diligence Roadmap" section at the end of the Deep Dive, suggesting exactly where a future technical auditor should look (e.g., "Audit the specific weights of the Layer 4 aggregator").
4. Inject Detail: Expand the document by integrating these new findings into verbose paragraphs.

12. PHASE 13: ADVANCED VISUAL ARGUMENTATION (SKILL DELEGATION)

CRITICAL MANDATE: Do NOT attempt to generate Excalidraw JSON manually. Use the activate_skill tool to invoke the excalidraw-diagram skill.

1. Invoke the Skill: Pass your detailed Visual Design Specification to the excalidraw-diagram skill.
2. Rendering: After the skill generates the .excalidraw files, use the internal renderer to convert them to PNGs:
   • Command: python3 ~/.agents/skills/excalidraw-diagram/references/render_excalidraw.py diligence_[startup]/diagrams/[filename].excalidraw

13. PHASE 13.1: VISUAL QUALITY ASSURANCE & REFINEMENT LOOP

Once diagrams are initially generated and rendered as PNGs, you MUST perform a Visual Audit:

1. Rendering Check: Inspect the PNG (or use file and ls -lh) to ensure it rendered correctly and contains text.
2. Logic Consistency: Does the diagram match the Isomorphism Principle? Does the structure make sense without reading labels?
3. Refinement Round: If the diagram is too crowded or lacks technical detail (failing the Education Test):
   • Update the Visual Design Specification with more granular box labels and specific technical primitives.
   • Re-delegate to the excalidraw-diagram skill.
   • Re-render the PNG.
4. Label Verification: Ensure the Two-Way Binding Rule is strictly followed in the new version.

14. PHASE 15: FINAL HOLISTIC SYNTHESIS & CONVICTION SCORING

Before finalizing the hand-off, perform a Holistic Global Analysis across all generated documents.

14.1 The Conviction Scorecard

Assign a score (1-10) to the following dimensions based on the entire output:

1. Technical Moat: (Is it a physics-based moat or a commodity wrapper?)
2. Founder Alpha: (Do they have the "Right to Win" and "Grit Signals"?)
3. Market Urgency: (Is this a "Painkiller" or a "Vitamin"?)
4. Capital Efficiency: (What is their ARR-to-Burn ratio?)
5. Regulatory/Platform Resilience: (Can they survive OpenAI or shifting laws?)

14.2 The "Red Team" Critique

Force the AI to argue against the investment for one paragraph. What is the single most likely reason this company fails?

14.3 Synthesis Statement

Construct a 3-paragraph final argument that weaves together the Architecture, PMF Level, and Capital Structure into a definitive Lead Partner verdict.

15. PHASE 16: EXCALIDRAW TECHNICAL REFERENCE & JSON PROTOCOL

To ensure diagrams are high-quality and correctly rendered, follow these strict JSON schema rules.

15.1 The Two-Way Binding Rule (CRITICAL)

Every text element inside a container MUST be bound correctly to prevent empty boxes.

1. Container (Rectangle/Ellipse/Diamond): Must have a boundElements array containing the text element's ID.
2. Text Element: Must have a containerId property pointing to the parent shape's ID.

15.2 Semantic Color Palette

Use colors to encode meaning from references/color-palette.md:

• AI/LLM: Fill #ddd6fe, Stroke #6d28d9
• Start/Trigger: Fill #fed7aa, Stroke #c2410c
• Success/End: Fill #a7f3d0, Stroke #047857
• Decision: Fill #fef3c7, Stroke #b45309
• Primary/Neutral: Fill #3b82f6, Stroke #1e3a5f

16. FINAL QUALITY ASSURANCE & HAND-OFF

19.1 Hand-off

"Diligence complete. Overview Memo, Deep Dive, and 30+ page Master Audit (with full feature teardowns, iteratively expanded details, and specialized visual diagrams) are in deliverables/."

End of VC Diligence Skill Masterclass V26.