gm-evaluate

What problem does it solve?

This skill prevents shipping a Godot tag that looks good in code but fails the playable closed loop, by independently evaluating the current tag’s final output quality and end-to-end behavior.

Core Features & Use Cases

• Maintains a single shared E2E suite: keeps e2e/ synchronized to the union of Tag Mechanics and Inherited Mechanics for the current tag.
• Enforces playable closed-loop gates: confirms the project builds cleanly, boots the correct entry scene, and reaches completion/fail/exit states through real gameplay.
• Validates gameplay quality with evidence: maps Playable Unit rows to passing runtime tests and records evidence in evaluation.json.
• Runs visual QA cross-checks per scene: captures screenshots, compares against references, and records verdicts with VQA logs.

Quick Start

Run the evaluation explicitly by sending the command /gm-evaluate for the current tag.