OSS Investment Scorecard provides a structured, weighted framework for Venture Capital funds to evaluate open-source AI projects, particularly during the AI acceleration cycle. Developed from practical experience and calibrated against real deals like vLLM/Inferact and Hugging Face, it aims to eliminate bias, especially for early-stage projects with scarce public data. The framework serves investors, founders, and analysts seeking objective project assessments.

How It Works

The framework employs a 5-dimension scoring system: Open-Source Ecosystem Health (25%), Team & Globalisation (20%), Technical Moat & Positioning (20%), Commercialisation & PMF (20%), and Capital Exit Path (15%). Version 1.2 introduces key improvements like a Mandatory Fact Sheet, Indirect Signal Inference for traction estimation, Project Age Calibration prioritizing velocity for younger projects, and Narrative Pivot Exemptions. Scores range from 5.5 to 10.0, with defined thresholds for recommendation, tracking, or passing, alongside six "One-Vote Vetoes" for automatic disqualification.

Quick Start & Requirements

The scorecard can be utilized in several ways:

• Option A (Claude AI): Download oss-investment-scorecard.skill and upload it to Claude.ai's Skills section for automated evaluation.
• Option B (Manual): Use the template/evaluation-template.md file to conduct a manual assessment.
• Option C (Any LLM Agent): Copy the content from SKILL.md (excluding the YAML header) into an LLM's system prompt or context window. No specific software prerequisites beyond access to an LLM agent (e.g., Claude, GPT-4, Gemini) are listed.

Highlighted Details

• V1.2 (March 2026) structural update focuses on bias elimination for Seed/Series A projects.
• Key V1.2 improvements include a Mandatory Fact Sheet, Indirect Signal Inference, Project Age Calibration, and Narrative Pivot Exemptions.
• Calibrated against real-world examples: vLLM/Inferact (8.9/10) and Hugging Face (8.5/10).
• Clear score thresholds (🟢 8.5–10.0, 🟡 7.0–8.4, 🟠 5.5–6.9, 🔴 < 5.5) and "One-Vote Vetoes" guide decision-making.

Maintenance & Community

The project is maintained by Lucy Chen, EIR at Zoo Capital ($2B+ AUM). Community submissions of project evaluations are encouraged via GitHub Issues or direct contact, contributing to a public record of assessed projects and facilitating connections between investors and founders.

Licensing & Compatibility

The framework is released under the MIT license, allowing free usage with attribution appreciated. It is compatible with commercial applications.

Limitations & Caveats

The framework is designed to mitigate inherent challenges in evaluating early-stage open-source projects where public data is often limited. The "One-Vote Vetoes" represent specific conditions that can override the calculated score, acting as critical decision gates.