gui-review

Reviews Unity MVVM GUI code through parallel subagent and mechanical verification gates.

Updated Jun 16, 2026
One-click install
npx skills add https://github.com/AvatarGanymede/dev-gui-plugin --skill gui-review-avatarganymede
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: gui-review
Source: https://github.com/AvatarGanymede/dev-gui-plugin/tree/main/skills/gui-review
Command: npx skills add https://github.com/AvatarGanymede/dev-gui-plugin --skill gui-review-avatarganymede

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Reviewing Unity MVVM GUI code (Panel.lua, View.cs, prefab bindings, config tables) requires both subjective quality judgment and deterministic machine checks, and doing them in one pass risks bias and missed compile errors. This Skill runs both lanes in parallel and merges results into a single verdict. ## Core Features & Use Cases - Parallel two-lane review: A fresh gui-reviewer subagent (Bias Guard, no prior context) judges taste dimensions while the orchestrator runs mechanical Type-A gates (C# compile, Lua syntax, prefab binding, generated-file export, config sync). - Merged CRITICAL gate with 6-state verdict: Produces GUI_REVIEW.md, GUI_VERDICT.json (PASS/WARN/FAIL/NOT_APPLICABLE/BLOCKED/ERROR), and a HUMAN_REVIEW.md checklist without blocking the pipeline. - Use Case: After generating a BagPanel with a new batch-sell button, run this Skill to verify the C# compiles, Lua passes LSP diagnostics, prefab bindings match, and an independent reviewer flags MVVM or lifecycle issues before acceptance. ## Quick Start Ask the AI to review the generated GUI code for a panel using the gui-review verification gate and produce the verdict JSON and human review checklist.

Frequently Asked Questions about gui-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review Unity MVVM GUI code automatically?▼

Run the gui-review stage, which spawns an independent gui-reviewer subagent for taste dimensions while the orchestrator runs mechanical checks like C# compilation and Lua syntax. Results merge into GUI_REVIEW.md and a GUI_VERDICT.json verdict.

What is a Bias Guard in AI code review?▼

Bias Guard means each review uses a fresh subagent with no prior thread context and no implementation narrative. The reviewer only sees the current code, GUI_PLAN.md, and shared references, so it judges quality purely from the code itself.

What checks does the mechanical Type-A gate run?▼

Type-A gates cover C# compilation (via unity-cli or Batch Mode), Lua syntax through LSP diagnostics, prefab node and binding verification, generated-file export diffs, and Excel-to-data config synchronization. Missing capabilities are marked BLOCKED, not failed.

What happens when the review finds CRITICAL issues?▼

If the merged CRITICAL count is greater than zero, the pipeline enters the gui-improve phase for iterative fixes, up to two rounds, rerunning both review lanes after each round. Zero CRITICALs skips improve and proceeds to knowledge learning.

Does the review pipeline stop for human approval?▼

No, there are no mid-pipeline checkpoints. Items needing human confirmation, including FAIL, BLOCKED, and ERROR verdicts, are collected into HUMAN_REVIEW.md while the pipeline continues running.