Analysis: Compare SkiaSharp and MAUI PR review agents, provide improvement recommendations

[Copilot]

Description of Change

Comprehensive analysis comparing SkiaSharp's PR review agent with .NET MAUI's agent to identify strengths, differences, and optimization opportunities for LLM AI models.

Analysis Documents Created:

• Executive Summary (265 lines) - Top 5 improvements with impact estimates and implementation roadmap
• Full Comparison (718 lines) - Detailed analysis of 10 similarities, 10 key differences, and domain-specific strengths
• Improvements Guide (1,006 lines) - Ready-to-implement code examples with 8 specific improvements prioritized by impact
• Quick Comparison Table (218 lines) - Side-by-side feature matrix with score cards and priority matrix

Key Findings:

Both agents share core strengths:

• Test-first philosophy requiring hands-on validation
• Unlimited time principle preventing rushed reviews
• Pause-and-ask pattern for handling blockers
• Structured workflow with feedback categorization (🔴 Critical / 🟡 Suggestions / 💡 Nitpicks)

Unique Strengths:

SkiaSharp excels at conceptual depth:

• Three-layer architecture verification (C++ → C API → C#)
• Memory management detail (3 pointer types: Raw/Owned/Ref-Counted)
• Deep understanding framework (explicit "think critically about WHY" guidance)
• Security focus and generated code handling

MAUI excels at actionable guidance:

• Explicit test app selection ("USE THIS / DON'T USE THAT") saving 20+ minutes
• Concrete instrumentation examples with copy-paste code
• Step-by-step comparison templates with exact commands
• Visual testing guidance for simulators

Top 5 Recommended Improvements for SkiaSharp:

1. Test Environment Selection (⭐⭐⭐⭐⭐) - Add explicit decision tree preventing 20-30 min waste on unnecessary native builds
2. Instrumentation Examples (⭐⭐⭐⭐) - Add copy-paste patterns for memory leak detection, disposal testing, stress testing
3. WITH/WITHOUT Comparison Template (⭐⭐⭐⭐) - Add structured commands and output formatting
4. Common Pitfalls with Visual Examples (⭐⭐⭐⭐) - Add before/after code showing wrong vs correct patterns
5. Testing Efficiency Decision Tree (⭐⭐⭐) - Add scope-based levels (5-10 min for small fixes, 30-60 min for C API changes)

Expected Impact: 20-40 minutes saved per review + significantly higher review quality

Implementation Effort: ~6 hours total across 3 phases

All analysis documents available in /tmp/ for review.

Bugs Fixed

None. Analysis-only task.

API Changes

None.

Behavioral Changes

None.

Required skia PR

None.

PR Checklist

• Has tests (N/A - analysis/documentation task)
• Rebased on top of main at time of PR
• Merged related skia PRs (N/A)
• Changes adhere to coding standard (N/A - no code changes)
• Updated documentation (analysis documents created)

Original prompt

This branch has a custom PR review agent.

The Maui project also has one https://raw.githubusercontent.com/dotnet/maui/refs/heads/main/.github/agents/pr-reviewer.md

Can you download the Maui one and compare.

Keep in mind they are different projects, but they both are designed to give excellent reviews on PRs.

Compare and indicate

• the good parts in both
• how they are similar
• how they differ
• how we can improve our one
• any suggestions or improvements we can make

This is only used by ai, so focus on making it work best with llm AI models.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.