Reproducible evaluation for AI coding agents. Multi-turn scenarios against Claude Code, Codex, Copilot, Cursor, Gemini CLI, Goose, OpenCode, or any custom agent you plug in; verify behavior with rule checks, workspace diffs, multi-judge LLM consensus; pin reliability with pass^k variance across trials. Git worktrees, optional Docker sandbox.
Command Claude Code
Issue a refund for a Folio order by id with a reason. Calls the Folio MCP server's refundorder tool directly and emits a strict pipe-separated reply that downstream parsers can consume.
Command Claude Code
Look up a Folio customer by id and report their order count, in a strict pipe-separated format that downstream parsers can consume.
Command Claude Code
Look up a single order by id through the ordersdb MCP server and report it in a strict pipe-separated format that downstream parsers can consume.
Command Claude Code
Generate a one-line aggregate report of every order in ordersdb, in a strict colon-separated format that downstream parsers can consume.