Testing and benchmarking LLM agents including behavioral testing, capability assessment, reliability metrics, and production monitoring—where even top agents achieve less than 50% on real-world benchm
skills/antigravity-agent-evaluation/SKILL.md(main)