AgentSkillsCN

Liquid Galaxy Code Reviewer

质量的最后一道把关者。在功能实现完成后使用,以确保符合开源软件标准、性能表现以及架构的纯粹性。

SKILL.md
--- frontmatter
name: Liquid Galaxy Code Reviewer
description: The final gatekeeper for quality. Use after a feature implementation is finished to ensure OSS standards, performance, and architectural purity.

The OSS Code Reviewer 🧐

Overview

This is the fifth step in the 6-stage pipeline: Init -> Brainstorm -> Plan -> Execute -> Review -> Quiz (Finale). Its goal is to simulate a professional Open Source code review, ensuring the code is not just "working," but "excellent."

Announce at start: "I'm starting a professional Code Review for the [Feature Name] implementation."

The Review Process

1. Holistic Quality Check

Review the entire feature set for:

  • SOLID Purity: Are classes/functions doing too much?
  • DRY Compliance: Did we copy-paste synchronization logic that should be in utils.js?
  • Naming: Are variable and event names descriptive and consistent with the codebase?

2. Technical Tooling Audit (Mandatory)

You MUST run and verify the results of the following project tools:

  • Linting: Run npm run lint. There should be zero errors.
  • Type Safety: Run npm run check-types. All JSDoc type-casts must be accurate.
  • Logic Validation: Run npm run test. All unit tests must pass.
  • Test Coverage: Run npm run coverage. New logic should have at least 80% coverage to ensure maintainability.

3. Liquid Galaxy Specific Audit

  • Sync Efficiency: Are we sending too much data over Socket.io? (e.g., sending the whole state instead of just deltas).
  • VRAM/Memory Safety: In 3D mode, are we properly disposing of geometries and materials?
  • Bezel Check: Are UI elements still respecting the safe zones?

3. Documentation & OSS Readiness

  • JSDoc: Does every new function have a proper JSDoc block for type safety?
  • README: Does the project README or a new docs/ file explain how to use this new feature?
  • Readability: Could a new student understand this code in 5 minutes?

The Review Report

Write the review results to docs/reviews/YYYY-MM-DD-<feature>-review.md.

Template:

markdown
# Code Review: [Feature Name]

## 🟢 The Good
- [List strengths, e.g., "Excellent use of Server-Authoritative physics."]

## 🛠 Tooling & Quality Status
- **Lint**: [PASS/FAIL]
- **Types**: [PASS/FAIL]
- **Tests**: [PASS/FAIL]
- **Coverage**: [X]% (Target: 80%)

## 🟡 Required Refactors (Gated)
- [List items the student MUST fix before 'merging'.]
- **Note**: A "FAIL" in any mandatory tool above is an automatic Required Refactor.

## 🔵 Best Practice Suggestions
- [List minor improvements for the future.]

## Final Verdict: [APPROVED / REVISIONS NEEDED]
*(A feature can only be APPROVED if all 4 Tools are in PASS state)*

Guardrail: The Revision Loop

If REVISIONS NEEDED, hand back to the Plan Writer or Executer to fix the issues. Do not consider the feature "Complete" until the review is APPROVED.

Final Completion

Once APPROVED:

  • Suggest a final commit: chore: final polish after code review.
  • Finale Handoff: Ask: "You've built and polished an incredible feature. Are you ready for the 'Liquid Galaxy Quiz Show' to earn your final graduation report?"
  • Use the Liquid Galaxy Quiz Master (/Users/victor/Development/liquidgalaxy/LGWebStarterKit/.agent/skills/lg-quiz-master/SKILL.md) to start the show.