Release5 min read

Introducing AI Quality Scores

How the platform evaluates every piece for code quality, test coverage, and reliability.

Morgan Liu @mliu·March 5, 2025

What is an AI Quality Score?

Every piece in the marketplace receives a score from 0 to 100. This score reflects the platform's automated assessment of code quality, test coverage, error handling, and documentation.

How It Works

When a piece is submitted, the AI analyzer runs a series of static and dynamic checks:

Static Analysis

TypeScript strictness and type coverage
Presence of input validation
Comprehensive error handling paths
No hardcoded secrets or credentials

Dynamic Analysis

The piece is deployed in a sandbox environment
A suite of test payloads is sent
Response times and error rates are measured
Edge cases are probed

Community Signals

Number of installs and reuses
User ratings and reports
Number of open issues and resolution time

What the Scores Mean

Score	Meaning
90–100	Production-grade. Well-tested, well-documented, actively maintained.
75–89	Reliable. May lack some edge case handling or documentation.
50–74	Functional. Works for common cases but may fail under unusual conditions.
Below 50	Experimental. Use with caution.

Improving Your Score

The quickest way to improve your score is to add input validation with Zod:

typescript
import { z } from "https://deno.land/x/zod@v3.23.8/mod.ts";

const payloadSchema = z.object({
  email: z.string().email(),
  template: z.string().min(1),
});

const payload = payloadSchema.parse(await req.json());

A validated piece scores an average of 18 points higher than one without validation.