---
name: benchmark
description: "Evaluate Fuse Studio against other Roblox skills with reproducible tasks and evidence"
disable-model-invocation: true
---

# Fuse Studio / benchmark

Task: $ARGUMENTS

Use `knowledge/tools/benchmark-methodology` and `benchmarks/README.md`. Keep model, project, assets, MCP access and test prompts identical. Report scores, pass/fail, time and tokens honestly; no unmeasured victory claims.

Common protocol: inspect existing project context first; respect project's coding/style conventions; use targeted local knowledge search only if needed; make minimal coherent edits; verify using available tools; report changed paths and unverified behavior. Follow main Fuse Studio server-authority, cleanup and performance standards.