VC Associate Agent Skills Pack (v0.3)

This is a skills bundle for simulating a top-decile venture associate. The skills are designed to force concrete outputs, not just checklists. Every skill answers:

• "Is this worth partner attention?"
• "What must be true?"
• "What's the fastest way to falsify it?"
• "What concrete artifact moves this forward today?"

Core principles

1. Force-ranked outputs - Top 10 companies, top 4 risks, top 3 next steps. No sprawling lists.
2. Must-be-true + fastest falsification test - Every thesis, every diligence, every sourcing decision.
3. Salesforce as mandatory gate - No "qualified" company exists without a record + next task + owner.
4. Steelman the bear case - Every memo includes the best argument against.
5. Time-boxed workflows with gates - Kill, continue, or escalate. No deals in limbo.
6. Measure your judgment - Brier scores, calibration tracking, learning rate over time.

Skills included

Atomic skills (individual capabilities)

Skill	Description
thesis-market-mapping/	Build market maps with force-ranked top 10, kill criteria, contrarian hypotheses
signal-sourcing-engine/	Sourcing loops as first-class objects with cadence metrics (10/3/1 weekly)
founder-meeting-runbook/	Meeting prep, structure, notes, and follow-up within 2 hours
diligence-risk-burndown/	Max 4 risks, fastest tests, anti-confirmation evidence required
investment-memo-writing/	First-line recommendation, steelman against, decision log
deal-winning-process/	Competitive round closing with value previews and daily action plans
term-sheet-triage/	Waterfall modeling, multi-investor stacks, participation caps
judgment-postmortem-calibration/	Brier scores, calibration curves, learning rate measurement
saas-metrics-diligence/	SaaS-specific metrics validation
portfolio-support-ops/	Portfolio support with SLAs, outcome tracking, and impact measurement
salesforce-crm-ops/	CRM operations with field mapping, dry-run mode, schema discovery

Compound skills (full workflows)

Skill	Description
dealflow-sourcing-to-memo/	End-to-end 3/5/10-day sprints with gates at each phase
diligence-sprint/	Time-boxed diligence with daily deliverables and kill gates
portfolio-helpdesk-sprint/	Time-boxed portfolio support with triage, SLAs, and metrics

What's new in v0.3

Salesforce scripts now production-ready:

• sf_upsert_lead.py - Dry-run mode, config-based field mapping, external ID upsert, custom VC fields
• sf_create_opportunity.py - Stage mapping, VC fields (pass reason, what-would-change)
• sf_create_task.py - Full config support
• sf_describe.py - Schema discovery for any org
• Config templates in salesforce-crm-ops/config/

Portfolio skills now have same rigor as dealflow:

• SLAs by request type (P0/P1/P2)
• Outcome measurement (success rate, impact rating)
• Help menu templates per company
• Sprint structure with daily gates

Judgment calibration now quantified:

• Brier score calculation for prediction accuracy
• Calibration curves (predicted % vs actual %)
• Learning rate tracking over quarters
• Archetype performance base rates
• Heuristics with attached probabilities

Term sheet analysis now handles complexity:

• Multi-investor liquidation stack modeling
• Full waterfall calculations at multiple exit values
• Participation cap analysis
• Side letter review checklist
• Competing term sheet comparison

Quick start

1. Clone this repo
2. Point your agent at this directory
3. For Salesforce integration:

   # Discover your org's schema
   python3 salesforce-crm-ops/scripts/sf_describe.py Lead Account Opportunity Task
   
   # Edit config files for your org
   vim salesforce-crm-ops/config/field_map.yaml
   vim salesforce-crm-ops/config/stages.yaml
   
   # Always dry-run first
   python3 salesforce-crm-ops/scripts/sf_upsert_lead.py --dry-run \
     --email founder@example.com --first Ada --last Lovelace --company ExampleAI

Validation

If you use the Agent Skills reference library:

skills-ref validate .

Philosophy

Top associates don't just run checklists. They constantly answer:

• "Is this worth partner attention?"
• "What must be true?"
• "What's the fastest way to falsify it?"
• "What concrete artifact moves this forward today?"

These skills force those outputs. No wandering. No "watching" without a recheck date. Every workflow ends with a clear decision.

And top associates measure their judgment. They track predictions, score them against outcomes, and improve their calibration over time. The judgment-postmortem skill now includes Brier scores and learning rate tracking so you can quantify whether you're getting better at picking.