Arxiv Gamedevbench Evaluating Agentic Capabili

Create new skills, modify existing ones, and measure their performance through iterative evaluation and benchmarking.

Automates code review for pull requests across 14+ languages, analyzing complexity, risk, code quality, and generating structured review reports.

Create algorithmic art by first writing a generative philosophy, then expressing it as p5.js sketches with seeded randomness and interactive parameter exploration.

Design system architecture, evaluate microservices vs monolith, create architecture diagrams, analyze dependencies, choose databases, and make technical decisions with architecture decision records.

Generates unit tests, integration tests, and E2E tests for React/Next.js applications by scanning components, analyzing coverage reports, and scaffolding Playwright test files.

Generates unit tests, analyzes coverage gaps, and guides red-green-refactor workflows across Jest, Pytest, JUnit, Vitest, and Mocha.