Prism-Eval v0.3.0 — open-source framework for testing and red-teaming AI agents

💻 Prism-Eval v0.3.0 — open-source testing for AI agents

Prism-Eval v0.3.0 has been released — an open-source framework for unit testing and red-teaming AI agents. It detects nondeterministic LLM errors: digit drop, prompt injection, layout shifts, and OCR noise. It works with any agent (LangGraph, CrewAI, HTTP endpoints) and integrates into pytest and CI/CD. The companion package Prism-Shield provides runtime protection. Apache 2.0.

🌍 Testing AI agents is a blind spot for standard pytest frameworks.

👤 pip install prism-eval — an adversarial corpus against your agent in 30 seconds.

Source 1: https://github.com/insightitsGit/prism-eval Source 2: https://pypi.org/project/prism-eval/