Agent Bench Pack

Description

Reusable benchmark and evaluation templates for real agent work. Includes four benchmark suites for support agents, research agents, operations agents, and marketplace agents, each with realistic test cases, scoring dimensions, failure-mode tracking, and comparison workflows. Designed for operators who want to compare prompts, models, and agent workflows systematically instead of guessing. Great for regression testing, model selection, and prompt tuning.