Deeptechnic — Technical Due Diligence Framework#

A structured system for conducting independent technology due diligence on hard-tech startup investments. The system follows a four-round iterative workflow and evaluates projects across 10 quality dimensions.

This is a meta-skill. It bundles sub-skills as references. Load the relevant reference for your current phase.

When to Use#

• Evaluating a hard-tech startup's BP/technical claims
• Conducting a first-pass technical assessment before expert interviews
• Reviewing team capability gaps and supply chain risks
• Generating structured due diligence reports

Not for: Financial due diligence, legal due diligence, or final investment decisions.

Quick Start#

text

1. Load references/skills-index.md to see available sub-skills
2. For current phase, load the corresponding reference skill(s):
   - 1st round:   load references/skills/deep-tech-dd-analysis + references/skills/anti-rationalization
   - 2nd round:   load references/skills/team-assessment + references/skills/supply-chain-audit
   - 3rd round:   load references/agents/tech-reviewer + agents/competition-analyst
   - Before output: load references/skills/quality-gate
   - After delivery: load references/skills/feedback-collection
3. Follow the Process steps in each loaded skill
4. Run quality gate before final output

The 10 Quality Dimensions#

Every assessment is scored against these dimensions (0.0-1.0):

#DimensionWeightDescription1claim_validation0.20Independent physics-based validation of tech claims2engineering_migration_gap0.15Analysis of lab-to-product engineering gaps3per_person_assessment0.15Per-member team capability matching4independent_competitor0.10BP-independent competitor discovery5supplier_level_analysis0.10Supplier-level supply chain audit6independent_analysis0.10Independent vs BP-derived analysis7team_gap_precision0.05Precision of structural gap identification8quantitative_score0.05Decimal-precision scoring (3.5/5 vs 4/5)9priority_levels0.05P0/P1/P2 question prioritization10overview_page0.05Executive summary on page one

Core Rules#

1. Every tech claim must have independent physics-based verification (formula + calculation)
2. Team assessment must name each member individually with capability matching
3. Competitor analysis must independently verify BP's claims (never trust BP's competitor list)
4. Supply chain analysis must name specific suppliers, not generic "domestic alternatives"
5. All output must pass the 10-point quality gate before delivery
6. User feedback must be captured and structured as training data after each delivery

Workflow#

Round 1 (First Pass)#

• Collect public info (papers, patents, news)
• Validate core tech claims (formula + calculation)
• Assess first team capability
• Generate question list for team

Round 2 (After Team Feedback)#

• Refresh model with team's answers
• Generate expert interview outline + supply chain questions

Round 3 (External Input)#

• Cross-validate multiple sources
• Compile interview records

Round 4 (Final)#

• Synthesize all findings
• Generate comprehensive DD report

Output Format#

Reports follow this structure:

1. Overview page (mandatory): Claim judgment table + core rating
2. Technical analysis
3. Team assessment
4. Supply chain analysis
5. Competitive landscape
6. Comprehensive evaluation
7. Open questions (P0/P1/P2)
8. Next round recommendations

Report file naming: Deeptechnic_<ProjectEngName>_<Phase>_<Date>.docx

Anti-Rationalization#

Claims you may be tempted to make — and why they're wrong:

| "I don't need to verify every claim" | Missed physical impossibility → failed investment | | "The BP's competitor list is enough" | BPs selectively omit domestic competitors | | "Team bio looks good, skip verification" | Cannot verify without public sources | | "This is just a quick pass" | First impressions lock in; do it right the first time |

Red Flags#

• All claim indicators are integers (no real test data)
• BP describes features without technical specifications
• Multiple extreme specs claimed simultaneously
• No third-party data anywhere in the BP
• BP competitor chart shows "all wins, no losses"
• "Confidential" used as reason to hide core data

Verification#

• All 6 core rules applied
• Quality gate passed before output
• Feedback recorded after delivery
• Training data appended to SkillOpt dataset