Uploaded June 2026 | Updated November 2026, 2 hours ago
The new theme engine is in. The design token system is designed. Now comes the part no one
wants to do manually — ripping out every legacy dynamic resource across every control and
replacing it with the new token system. In this episode, I use AI agents to do the grunt work
so I don't have to.
But this isn't vibe coding. The key is the prompt. I walk through exactly how I primed and
instructed the AI — surgically, precisely — so it executed the migration without going off the
rails. The AI even surprised me with a markup extension solution I hadn't thought of. This is
Dev + AI working as it should: human in full control, AI doing the heavy lifting.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📌 WHAT YOU'LL LEARN
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
• How to structure an AI prompt to do precise, surgical code migration — not vague generation
• Why priming your AI agent with context before the task changes everything
• What happened when a local AI tried to grant itself admin access — and why security layers matter
• How the AI handled replacing all legacy dynamic resources with the new design token system
• Iterative prompting: how to guide the AI through questions and scope changes in real-time
• The unexpected markup extension solution the AI came up with — and why it was the right call
• How to use value converters for asymmetric padding/margin scenarios in the token system
• What a one-hour AI-assisted dev cycle looks like vs days of manual migration
• When to let AI make decisions vs when to pull it back — staying in control throughout
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⏱ TIMESTAMPS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[0:00] – Introduction — the new theme engine & design token system recap
[0:47] – What the design token system gives us: surface, on-surface, semantic naming
[1:33] – The problem: manual migration across every control would take days
[2:25] – The solution: AI agents as a surgical migration tool
[4:45] – The AI agent tool — chat view, Kanban board, and security features
[6:01] – Crafting the initial migration prompt — exactly what I told the AI
[9:26] – ⚠️ The rogue AI moment — local model tries to get admin access
[12:05] – AI starts executing, asks clarifying questions, user provides guidance
[16:19] – First pass complete: button spinner migrated, user reviews accuracy
[18:32] – Expanding the task: full reskin across all controls + delete orphaned files
[27:49] – Second pass done — AI raises questions about remaining base resources
[28:51] – Brainstorming: tokenisation approach vs value converters for complex padding
[36:04] – AI proposes final approach, user approves parallel execution
[44:51] – Final outcome: clean minimal theme structure + the markup extension surprise
[47:20] – Reflections on AI-assisted development — what this changes going forward
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔗 LINKS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
▶ Part 1 – Setup: youtube.com/watch?v=rUKlQnjU0tc
▶ Part 2 – Buttons: youtube.com/watch?v=oKjk6pyyARI
▶ Part 3 – Button Spinner Anatomy: youtube.com/watch?v=WMaTbXkvm9g
▶ Part 4 – Designing a Theme System: youtube.com/watch?v=I5halHoNfH0
▶ Part 5 – Checkbox Anatomy: youtube.com/watch?v=uIimnwjITT0
▶ Part 6 – Checkbox Styling: youtube.com/watch?v=uv1bNrfCjtY
▶ Part 7 – Theme Engine From Scratch: youtube.com/watch?v=NXz8hwITg5U
▶ Part 8 – Design Token System: youtube.com/watch?v=QMVdxRZqaU8
📺 Full Series Playlist: youtube.com/playlist?list=PLrW43fNmjaQVE2vDBeWD8QYIm06w6bKyO
💾 Source Code: github.com/angelsix/youtube/tree/develop/AvaloniaThemeLab
🌐 AngelSix: angelsix.com
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📺 SERIES: Avalonia UI Theming – Hand Coded From Scratch
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
#avalonia #avaloniaui #xaml #csharp #dotnet #aiagent #aitools #softwaretutorial #coding
The new theme engine is in. The design token system is designed. Now comes the part no one
wants to do manually — ripping out every legacy dynamic resource across every control and
replacing it with the new token system. In this episode, I use AI agents to do the grunt work
so I don't have to.
But this isn't vibe coding. The key is the prompt. I walk through exactly how I primed and
instructed the AI — surgically, precisely — so it executed the migration without going off the
rails. The AI even surprised me with a markup extension solution I hadn't thought of. This is
Dev + AI working as it should: human in full control, AI doing the heavy lifting.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📌 WHAT YOU'LL LEARN
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
• How to structure an AI prompt to do precise, surgical code migration — not vague generation
• Why priming your AI agent with context before the task changes everything
• What happened when a local AI tried to grant itself admin access — and why security layers matter
• How the AI handled replacing all legacy dynamic resources with the new design token system
• Iterative prompting: how to guide the AI through questions and scope changes in real-time
• The unexpected markup extension solution the AI came up with — and why it was the right call
• How to use value converters for asymmetric padding/margin scenarios in the token system
• What a one-hour AI-assisted dev cycle looks like vs days of manual migration
• When to let AI make decisions vs when to pull it back — staying in control throughout
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⏱ TIMESTAMPS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[0:00] – Introduction — the new theme engine & design token system recap
[0:47] – What the design token system gives us: surface, on-surface, semantic naming
[1:33] – The problem: manual migration across every control would take days
[2:25] – The solution: AI agents as a surgical migration tool
[4:45] – The AI agent tool — chat view, Kanban board, and security features
[6:01] – Crafting the initial migration prompt — exactly what I told the AI
[9:26] – ⚠️ The rogue AI moment — local model tries to get admin access
[12:05] – AI starts executing, asks clarifying questions, user provides guidance
[16:19] – First pass complete: button spinner migrated, user reviews accuracy
[18:32] – Expanding the task: full reskin across all controls + delete orphaned files
[27:49] – Second pass done — AI raises questions about remaining base resources
[28:51] – Brainstorming: tokenisation approach vs value converters for complex padding
[36:04] – AI proposes final approach, user approves parallel execution
[44:51] – Final outcome: clean minimal theme structure + the markup extension surprise
[47:20] – Reflections on AI-assisted development — what this changes going forward
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔗 LINKS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
▶ Part 1 – Setup: youtube.com/watch?v=rUKlQnjU0tc
▶ Part 2 – Buttons: youtube.com/watch?v=oKjk6pyyARI
▶ Part 3 – Button Spinner Anatomy: youtube.com/watch?v=WMaTbXkvm9g
▶ Part 4 – Designing a Theme System: youtube.com/watch?v=I5halHoNfH0
▶ Part 5 – Checkbox Anatomy: youtube.com/watch?v=uIimnwjITT0
▶ Part 6 – Checkbox Styling: youtube.com/watch?v=uv1bNrfCjtY
▶ Part 7 – Theme Engine From Scratch: youtube.com/watch?v=NXz8hwITg5U
▶ Part 8 – Design Token System: youtube.com/watch?v=QMVdxRZqaU8
📺 Full Series Playlist: youtube.com/playlist?list=PLrW43fNmjaQVE2vDBeWD8QYIm06w6bKyO
💾 Source Code: github.com/angelsix/youtube/tree/develop/AvaloniaThemeLab
🌐 AngelSix: angelsix.com
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📺 SERIES: Avalonia UI Theming – Hand Coded From Scratch
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
#avalonia #avaloniaui #xaml #csharp #dotnet #aiagent #aitools #softwaretutorial #coding










