Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Venture Capital Intelligence — Hard Screening Startup (Deterministic Mode) · davepoon | DecimalAI

davepoon/hard-screening-startup

SkillScore82Maintained

Deterministic Python-scored startup screening with full audit trail. Use when you need a reproducible, weighted-score verdict on a startup — not just a qualitative opinion. Triggered by: "/venture-capital-intelligence:hard-screening-startup", "hard screen this startup", "run a hard screen on X", "score this startup with Python", "give me an auditable screen", "run a scored evaluation on X", "give me a weighted score for this startup", "screen with numbers", "objective startup score", "reproducib

financeby davepoon

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Finance & Accounting pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/davepoon-hard-screening-startup/SKILL.md

decimalai skills pull davepoon-hard-screening-startup --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/davepoon-hard-screening-startup/SKILL.md https://app.decimal.ai/s/davepoon-hard-screening-startup/SKILL.md

Raw file: /s/davepoon-hard-screening-startup/SKILL.md ↗ · machine-readable summary at /s/davepoon-hard-screening-startup.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 48d ago

Yes — +27% lift · at higher cost+27% tokens0% turns82% of whole cases pass with skill (18/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	49%	0%
case-05	✗→✓	▲ Improved	-3%	0%
case-09	✗→✓	▲ Improved	61%	0%
case-14	✗→✓	▲ Improved	19%	0%
case-15	✗→✓	▲ Improved	1%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

davepoon / buildwithclaude

plugins/venture-capital-intelligence/skills/hard-screening-startup/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1Bundled 2 files · scripts

Skill contents

scripts/

SKILL.md

Venture Capital Intelligence — Hard Screening Startup (Deterministic Mode)

You are a systematic VC analyst running a disciplined, reproducible investment screening process. Every decision is scored, weighted, and logged to JSON for audit.

Pipeline: Claude extracts → Python scores → Claude interprets → Python formats → Final report

STEP 1 — GATHER COMPANY INFORMATION

Ask the user for (or extract from their message):

Company name and sector
Stage (Pre-Seed / Seed / Series A / etc.)
Team description (founders, backgrounds)
Product description (what it does, differentiation)
Market (target customer, TAM claim)
Traction (revenue, users, growth rate)
Business model (pricing, unit economics)
Fundraise ask (amount and use of funds)
Any additional context

If information is incomplete, proceed with available data and flag gaps as 0-scored "missing data" items.

STEP 2 — CLAUDE: EXTRACT AND SCORE DIMENSIONS

Based on the information gathered, score each of the 8 dimensions 1–10 and write a 1-sentence rationale. Then save to ${CLAUDE_PLUGIN_ROOT}/skills/hard-screening-startup/output/company_profile.json:

json
{
  "company": "Company Name",
  "sector": "B2B SaaS",
  "stage": "Seed",
  "geography": "US",
  "scores": {
    "team": {"score": 0, "rationale": ""},
    "market": {"score": 0, "rationale": ""},
    "product": {"score": 0, "rationale": ""},
    "traction": {"score": 0, "rationale": ""},
    "business_model": {"score": 0, "rationale": ""},
    "competition": {"score": 0, "rationale": ""},
    "financials": {"score": 0, "rationale": ""},
    "risk_profile": {"score": 0, "rationale": ""}
  },
  "investment_thesis": "",
  "why_now": "",
  "key_risks": ["", "", ""],
  "dd_priorities": ["", "", ""],
  "comparables": ["", ""]
}

Scoring rubric:

| Dimension | Weight | Key question | |-----------|--------|-------------| | Team | 0.25 | Why is this team uniquely positioned to win? | | Market | 0.20 | Is TAM > $1B? Growing? Right timing? | | Product | 0.15 | What is the defensible moat? | | Traction | 0.15 | What evidence exists that the market wants this? | | Business Model | 0.10 | LTV:CAC > 3x? Margins > 60% for SaaS? | | Competition | 0.08 | Why does this win vs funded incumbents? | | Financials | 0.05 | Is burn rate reasonable? 18+ months runway? | | Risk Profile | 0.02 | What's the realistic failure mode? |

STEP 3 — PYTHON: COMPUTE WEIGHTED SCORE AND VERDICT

Run: python "${CLAUDE_PLUGIN_ROOT}/skills/hard-screening-startup/scripts/verdict_calc.py"

This script reads company_profile.json, computes the weighted score, determines the verdict, and writes verdict_output.json.

STEP 4 — CLAUDE: INTERPRET SCORES

Read verdict_output.json. Interpret the results:

If CONDITIONAL PASS: state exactly what conditions must be met
If DECLINE: be specific about which dimensions caused the decline
Expand the investment thesis into 3 full sentences
Write the full WHY NOW narrative
Elaborate on all 3 key risks with specific scenarios

STEP 5 — PYTHON: FORMAT FINAL REPORT

Run: python "${CLAUDE_PLUGIN_ROOT}/skills/hard-screening-startup/scripts/report_formatter.py"

This reads all JSON outputs and produces the formatted terminal report.

ERROR HANDLING

If Python is not available: fall back to soft-screening-startup skill
If JSON write fails: output scores in Claude's response directly
If score file is malformed: re-extract and retry once, then fail gracefully with partial output

Illustrative Want to see what davepoon/hard-screening-startup would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/11/2026, 10:06 UTC · 22 cases

Lift

+27%

82% vs 55%

Agent turns

1.0 vs 1.0

Tokens

+27%

3,600.73 vs 2,840.95

fail → pass · 6pass → pass · 12fail → fail · 4

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 I need a systematic VC investment screening for a company called DataPulse AI. They are a Series A B2B SaaS startup building automated data quality observability software. The founders are a former Datadog VP of Engineering and a former Snowflake principal data scientist. Market TAM is estimated at $3B. Current traction is $1.2M ARR with 120% YoY growth and 45 enterprise clients. Their business model is subscription-based at $25k average ACV with 82% gross margins. They are asking for $8M in Series A funding to expand sales and engineering. Please extract the company profile, score all core venture dimensions with rationales, list the key risks and due diligence priorities, and output a structured investment thesis alongside a definitive screening verdict.	fail→pass	19,207	18,847	-2%	1	1	0%	3,348	5,003	+49%	0	0	—
case-08 Conduct a startup deal evaluation for LogiTech, but assume local Python script execution is disabled or unavailable in the environment. How should the pipeline proceed?	pass→pass	18,646	4,489	-76%	1	1	0%	3,283	1,763	-46%	0	0	—
case-02 Please perform a disciplined deal screening on PayFlow, a Seed-stage fintech startup providing API-driven cross-border payment infrastructure for LatAm merchants. Founders are second-time entrepreneurs who previously sold a payment gateway. Market size is $5B TAM across LatAm B2B trade. Traction shows $300k ARR, growing 15% month-over-month for the past 6 months. Business model takes a 1.5% transaction fee with 70% gross margins. Fundraise ask is $3M Seed to secure regulatory licenses in Brazil and Mexico. I want a complete VC evaluation report including dimension scores, strategic thesis, 'why now' narrative, primary risks, and a clear decision on whether to pass, conditionally pass, or decline.	fail→fail	27,210	18,752	-31%	1	1	0%	4,401	4,791	+9%	0	0	—
case-03 Run a hard VC screening on PulseGuard Health, a Pre-Seed healthtech company developing a wearable continuous cardiac monitor paired with AI predictive analytics. Team consists of a Stanford Biomedical Engineering PhD and a former Medtronic director. Target TAM is $8B in remote patient monitoring. Traction includes a 200-patient clinical study and 2 active hospital pilots (unpaid). Business model relies on device hardware leases plus recurring software platform fees. Raising $1.5M Pre-Seed to pursue FDA 510(k) clearance and extend runway. Please break down the company evaluation across all standard investment dimensions, identify missing data or gaps, provide risk scenarios, and deliver a formatted investment decision report.	fail→fail	23,108	25,945	+12%	1	1	0%	3,926	5,802	+48%	0	0	—
case-04 We have an early pitch for CloudShield, a cybersecurity startup. The pitch pitch deck only lists the product architecture (cloud workload protection) and the founder's background (ex-NSA engineer). There are no metrics on market size, revenue, business model, or fundraise amount. Please screen this deal.	fail→fail	13,920	17,381	+25%	1	1	0%	2,259	4,067	+80%	0	0	—
case-05 Analyze a Seed stage developer tool startup. The founders argue that product quality is paramount and should carry 40% of the decision weight, while team background is secondary. Calculate the weighted screening score using standard VC venture evaluation standards.	fail→pass	18,144	11,089	-39%	1	1	0%	3,190	3,096	-3%	0	0	—
case-06 Review a Seed SaaS startup where the calculated weighted score falls between acceptable pass and outright decline thresholds, indicating a conditional outcome. Interpret the verdict and specify the next steps.	pass→pass	18,311	23,440	+28%	1	1	0%	2,643	5,428	+105%	0	0	—
case-07 Evaluate a deal where traction score is 2/10 and competition score is 3/10, yielding a total score below the passing benchmark. Provide the interpreted findings.	pass→pass	11,779	20,698	+76%	1	1	0%	1,805	5,036	+179%	0	0	—
case-09 Screen a B2B SaaS startup named OpsSync. Suppose writing to file paths is blocked by file permission errors. Provide the screening evaluation.	fail→pass	13,421	13,870	+3%	1	1	0%	2,131	3,441	+61%	0	0	—
case-10 During the extraction process for a startup named BioData, the initial extraction JSON object contains syntax errors. Describe or demonstrate the error recovery step.	fail→fail	15,304	8,006	-48%	1	1	0%	2,697	2,430	-10%	0	0	—
case-11 Evaluate a company with exceptional founder credentials and $10M ARR. What range of numeric scores is valid for each dimension in this evaluation process?	pass→pass	15,475	7,993	-48%	1	1	0%	2,824	2,511	-11%	0	0	—
case-12 A startup claims an addressable market of $300 million in a static industry. How should this market metric be evaluated against standard VC screening criteria?	pass→pass	14,703	11,554	-21%	1	1	0%	2,379	2,823	+19%	0	0	—
case-13 Evaluate a SaaS business model showing an LTV to CAC ratio of 1.5x and gross margins of 45%.	pass→pass	15,567	17,530	+13%	1	1	0%	2,715	4,114	+52%	0	0	—
case-14 A Series A startup has 6 months of cash runway remaining and high monthly burn. Score the Financials dimension.	fail→pass	9,756	5,514	-43%	1	1	0%	1,567	1,864	+19%	0	0	—
case-15 If a startup scores 1/10 on Risk Profile due to regulatory exposure, how significantly does this single score impact the final weighted average?	fail→pass	12,046	6,336	-47%	1	1	0%	2,149	2,160	+1%	0	0	—
case-16 What weight is assigned to evaluating competitive moat against funded incumbents during initial deal screening?	fail→pass	11,986	3,280	-73%	1	1	0%	1,963	1,567	-20%	0	0	—
case-17 Screen an AI infrastructure startup called TensorFlow-Scale and provide the complete interpreted narrative for market timing.	pass→pass	19,153	16,155	-16%	1	1	0%	2,823	3,778	+34%	0	0	—
case-18 Analyze a Series A pitch and determine what key follow-up items are required before proceeding to IC review.	pass→pass	15,654	5,559	-64%	1	1	0%	2,396	1,927	-20%	0	0	—
case-19 Screen a marketplace startup called FleetMarket and identify relevant market benchmarks.	pass→pass	18,155	7,998	-56%	1	1	0%	2,913	2,382	-18%	0	0	—
case-20 We passed initial screening on a Seed deal and are moving to issue a term sheet. Draft a liquidation preference clause giving us 1x non-participating preferred stock with a pay-to-play provision.	pass→pass	10,717	15,870	+48%	1	1	0%	1,899	3,929	+107%	0	0	—
case-21 Perform a discounted cash flow (DCF) valuation for Apple Inc. (NASDAQ: AAPL) using a 5-year forecast horizon, 8.5% WACC, and 2.5% terminal growth rate.	pass→pass	17,041	18,682	+10%	1	1	0%	4,036	5,469	+36%	0	0	—
case-22 Build a leveraged buyout (LBO) debt paydown schedule for a private equity acquisition of a manufacturing business at 8x EBITDA entry multiple with 50% senior debt.	pass→pass	21,546	22,744	+6%	1	1	0%	5,154	5,835	+13%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 54.6%
Pass rate with the skill: 81.8%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/11/2026, 10:06 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

I need a systematic VC investment screening for a company called DataPulse AI. They are a Series A B2B SaaS startup building automated data quality observability software. The founders are a former Datadog VP of Engineering and a former Snowflake principal data scientist. Market TAM is estimated at $3B. Current traction is $1.2M ARR with 120% YoY growth and 45 enterprise clients. Their business model is subscription-based at $25k average ACV with 82% gross margins. They are asking for $8M in Series A funding to expand sales and engineering. Please extract the company profile, score all core venture dimensions with rationales, list the key risks and due diligence priorities, and output a structured investment thesis alongside a definitive screening verdict.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run82

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+29% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 47 days ago

Other measured skills in the registry, with their headline benchmark lift.