Context

MoonBit is a Rust/Go-influenced language with structural types, pattern matching, and a focus on WebAssembly compilation. Like Vera and Aver, it has relatively little training data in LLM corpora, making it an interesting third data point in the zero-training-data comparison.

From the languages briefing: MoonBit emphasises performance, developer experience, and WASM-first compilation. It has algebraic data types, pattern matching, and a trait system — closer to Rust than to Vera's effect-handler model.

What's needed

Following the same pattern as the Aver integration (PR #48):

• 50 canonical MoonBit solutions in solutions/moonbit/
• MoonBit baseline runner in baseline_runner.py
• MoonBit prompt builder in prompts.py (loads MoonBit's language spec)
• MoonBit evaluation in runner.py
• CLI integration (--language moonbit)
• Unit tests for MoonBit-specific helpers
• .coderabbit.yaml exclusions for !**/*.mbt and !solutions/moonbit/**

The description_neutral field (added in PR #48) means MoonBit prompts will use language-neutral problem descriptions out of the box.

Relates to

• #48 — Aver support (same integration pattern)
• #21 — Go support (another comparison language)
• #24 — Multi-model evaluation