# AI Agent Evaluation Platform

Evaluate tool-using AI agents on real tasks with isolated sandboxes, scorecards, replay evidence, and CI regression gates.

Source: https://www.agentclash.dev/platform/agent-evaluation
Markdown export: https://www.agentclash.dev/md/platform/agent-evaluation

## Evaluation workflow

- Package tasks, inputs, tools, policies, and evaluators as challenge packs.
- Race candidate agents under the same constraints.
- Inspect scorecards and replay evidence.
- Promote failures and gate future releases.