vs-ir-eval is a public framework for reviewing startup IR decks, pitch materials, and business plans from a VentureSquare-style mentoring perspective.
It goes beyond simple summarization. The skill asks practical early-stage questions such as:
- Are the founders meeting real customers?
- Is the business hard for others to copy?
- Is the team chasing unfinished technology without enough customer evidence?
- Can the company reach paid usage, operating leverage, and a credible exit path?
The 2026-05 version also uses public Sequoia, Y Combinator, a16z, Bessemer, marketplace, and SaaS evaluation lenses as supporting references. This repository is for public use, so it does not replace real investment decisions, investment advice, confidential investment committee standards, or due diligence.
Before using this in a public demo or sharing it externally, read Public Use Notice.
The English public version is the default at the repository root. The Korean version is kept alongside it for Korean founders and reviewers:
The demo web app also supports Korean and English output. Select the preferred output language before running a review.
Give an OpenClaw agent or compatible prompt runtime a business plan, pitch deck text, product summary, or idea brief and ask:
"Evaluate this business plan with the vs-ir-eval skill."
"Analyze this startup in a VentureSquare-style review."
The recommended default is Coaching mode. For public demos and founder self-checks, it is safer to focus on questions, missing evidence, and next steps instead of showing scores first.
Use Screening mode when you need a compact investor-style pre-review memo. Use Full report mode only when a longer diagnostic report is explicitly needed.
This skill is a mentoring and preparation aid. Do not outsource investment judgment to AI.
- It must not be used to produce investment advice, suitability judgments, investment commitments, pass/fail notifications, or official VentureSquare investment decisions.
- Public demos must not receive confidential IR materials, personal information, contracts, source financial documents, term sheets, cap tables, shareholder lists, or trade secrets.
- Reviewers must verify outputs through source materials, founder meetings, customer validation, financial/legal/technical due diligence, and human judgment before using them in any decision process.
- AI outputs can vary by model state, prompt interpretation, runtime conditions, and traffic load.
- Numeric scores are secondary ordering aids. Public-facing judgment should use grade bands and traceable evidence.
- Founder claims and externally verified facts must be separated.
- If live browsing is unavailable, the output must mark external verification as
Not performed (no browsing tool)orUnverified.
- Executive Summary: one-line assessment, review opinion, rationale, and the largest unverified assumption
- VS Capability Radar and Evidence Trace: stage classification, stage-based weighting, Team/Market/Moat/Scale/Strategy scores, and evidence gaps
- TIPS / LIPS Fit Diagnosis: technology-startup or local/lifestyle program fit, suggested project title, and required evidence
- VentureSquare-Style Upside and Downside Factors: cost structure, funding ability, time to monetization, dependency risks, and operating-profit path
- Global Investor Framework Cross-Check: Sequoia, YC, a16z/Bessemer, marketplace, and SaaS lenses where applicable
- Market, Competitor, and External Fact Check: separate founder claims, verified facts, and unverified items
- Comparable Cases and Valuation Readiness: comparable candidates and valuation assumptions only when data quality supports them
- VCS Investor/Fund Search Guide: official VCS search guidance and, only when verified, public-information-based sample outreach candidates
- Deep Dive: strengths and red flags
- Investment Review Checklist: numbers, documents, and pass conditions to verify next
- Public Mentoring Advice: pressure questions, missing proof, and next actions for the founder
Unless the user explicitly forbids web search, the skill should perform current internet research before producing a report.
If the runtime does not provide live web search or browsing tools:
- Use only sources and URLs supplied by the user.
- Do not imply that external verification was performed.
- Mark external verification as
Not performed (no browsing tool)orUnverified. - Turn market, valuation, and VCS sections into search guidance and verification checklists.
Research should prioritize official company materials, filings, DART/SEC/exchange data, investor or acquirer announcements, government/public statistics, patent/clinical/regulatory databases, credible media, and reputable investment databases.
| Weighted score | Public grade | Meaning |
|---|---|---|
| 90-100 | Excellent | Strong evidence across team, market, moat, scale, and strategic rationale |
| 75-89 | Strong | Worth deeper review, with important verification items remaining |
| 60-74 | Moderate | Useful for mentoring or observation, but evidence must improve before investment review |
| 45-59 | Weak | One or more core assumptions, evidence areas, or business-model elements need major repair |
| 0-44 | Needs Work | Too many unverified assumptions or fatal risks; use primarily for coaching |
- Coaching mode: no total score or radar-first framing; focus on pressure questions, missing evidence, and next actions
- Screening mode: one-page investor memo with grade, evidence, risks, and follow-up checks
- Full report mode: full diagnostic report with all sections
The public web app defaults to Coaching mode because founders can misread score-first outputs as pass/fail judgments.
| Stage | Team | Market | Moat | Scale/Exit | Strategy/TIPS |
|---|---|---|---|---|---|
| Pre-incorporation | 35% | 25% | 15% | 10% | 15% |
| Under 3 years | 25% | 25% | 20% | 15% | 15% |
| Under 5 years | 20% | 20% | 20% | 25% | 15% |
| Under 7 years | 15% | 20% | 20% | 30% | 15% |
| 7+ years | 10% | 15% | 20% | 40% | 15% |
Key evidence by stage:
- Pre-incorporation: founder-problem fit, customer interviews, MVP plan, initial funding plan
- Under 3 years: MVP/PoC, first paid customers, usage logs, technical/regulatory/IP validation
- Under 5 years: revenue, retention, CAC/LTV, gross margin, customer references
- Under 7 years: repeat revenue, sales pipeline, break-even path, follow-on funding or exit feasibility
- 7+ years: financial statements, operating-profit path, market share, IPO/M&A comparables, governance risks
Upside factors:
- Simple cost structure and realistic founder/team financing capacity
- Short path from product completion to paid customer usage
- Potential for operating-profit conversion through automation, cost reduction, or unit-economics improvement
- Repeatable revenue supported by systems and process, not only founder heroics
Downside factors:
- Founder or core team lacks relevant experience, domain knowledge, or execution history
- Heavy cost structure that may remain unprofitable even after multiple funding rounds
- High dependency on one customer, partner, platform, institution, grant, or manual project work
- Revenue may exist, but the path to operating profit is unclear
The final opinion should explain either why a downside is tolerable or why an upside is still insufficient.
- Sequoia Capital: purpose, problem, solution, why now, market, competition, product, business model, team, financials, and milestones
- Y Combinator: making something people want, early user love, launch discipline, user conversations, and doing things that do not scale
- a16z: revenue quality, CAC/LTV, gross margin, churn, retention, burn, engagement, and metric discipline
- a16z Marketplace: GMV, take rate, liquidity, fill/match rate, repeat usage, and disintermediation risk
- Bessemer: SaaS/cloud ARR/MRR quality, CAC payback, net revenue retention, gross margin, and burn multiple
Reference links:
- https://sequoiacap.com/article/writing-a-business-plan/
- https://www.ycombinator.com/blog/ycs-essential-startup-advice/
- https://www.ycombinator.com/blog/how-not-to-fail/
- https://a16z.com/16-startup-metrics/
- https://a16z.com/the-marketplace-glossary/
- https://www.bvp.com/atlas/10-laws-of-cloud
- https://www.vcs.go.kr/web/portal/investor/list
- https://www.vcs.go.kr/web/portal/rsh/list
Copy this repository into your OpenClaw skills directory.
cd ~/.openclaw/workspace/skills
git clone https://github.com/VScommonBot/vs-ir-eval.gitweb-app/index.php is a single-file PHP demo app. Do not place API keys in source code. Inject them only through server environment variables.
export OPENAI_API_KEY="<your-openai-api-key>"
export OPENAI_MODEL="<your-preferred-chat-model>"
php -S 127.0.0.1:8080 -t web-appPublic deployment checklist:
- Tell users that submitted text is sent to the OpenAI API and require consent.
- Warn users not to submit confidential IR materials, personal information, contracts, source financials, term sheets, cap tables, shareholder lists, or trade secrets.
- Keep Coaching mode as the public default.
- Keep Korean and English outputs available. Do not remove the Korean prompt/doc files when updating the English public version.
- Sanitize Markdown before rendering it as HTML. The demo app uses DOMPurify.
- Add request-size limits, rate limiting, monitoring, and logging appropriate for your server.
- Pin CDN assets with integrity checks or self-host them before production deployment.
- The demo app does not provide live browsing tools. When external verification is needed, connect a separate search/browsing tool or provide verified sources in the input.
This framework reflects a public VentureSquare-style mentoring view:
- Conditions for growth: a timely market, appropriate technology, founders who meet real customers, a defensible reason why this team can win, and a team that can persist.
- Conditions for failure: unclear customer definition, obsession with unfinished technology, uncontrollable regulatory risk, weak leadership, and no path to paid usage or operating leverage.
- Investor lens: exit feasibility, M&A/global potential, public-support fit such as TIPS/LIPS, and evidence quality.