"author:itsryanarmani" matching MCP servers:

Enables listing and retrieving tasks from a benchmark for rule-conflict scenarios, scoring model answers, running evaluations, and checking data provenance through MCP tools. It supports offline mock runs and OpenRouter-backed model evaluations with deterministic assertion grading.