Skill: AI Collaboration Workflow

Last Verified: 2026-01-23 Applicable SDK: Android 14+ (API 34+) Dependencies: None

Purpose

This skill provides proven techniques for effective collaboration with AI coding assistants, maximizing productivity and reducing misunderstandings.

When to Use This Skill

•Starting a new project with AI assistance
•Experiencing misunderstandings or incorrect implementations
•Want to improve AI collaboration efficiency

Principle 1: Config-First Project Setup

Before writing features, establish the foundation:

•Create colors.xml: Define semantic colors (bg_primary, text_secondary)
•Create themes.xml: Configure Day/Night mode with explicit windowBackground
•Create utility classes: Utils.dp2px, HapticManager, SoundManager

Why: Extracting colors and dimensions later wastes significant refactoring time.

Principle 2: Clear Requirement Description

The Golden Formula

[Location/Area] + [Current Behavior] + [Reference/Sensory Description] = Precise Execution

Examples

❌ Vague	✅ Clear
"Fix that clock thing"	"The clock digits in `FlipCard.kt` are not centered. They should align to the visual ink center, not the font metrics center."
"Add burn-in protection"	"Add OLED burn-in protection: shift X/Y by 1px every minute, max 4px range, use ValueAnimator, reset to center on app restart."
"The line is too long"	"The middle divider line in White Widget should match the width of the card's narrowest point (the 'waist')."

Principle 3: Provide Context, Not Just Commands

Bad

"Edit MainActivity.kt"

Good

"I'm building the Settings page. I need to add a toggle for OLED protection. The toggle should:

•
Default to OFF

•
Persist to SharedPreferences

•
Immediately apply when changed"

Principle 4: State Constraints Explicitly

Always mention:

•Library restrictions: "Don't add new dependencies"
•API level: "Must work on API 26+"
•Performance: "Must maintain 60fps during animation"
•Locked code: "Don't modify the centering formula in FlipCard.kt"

Principle 5: Describe Sensory, Not Technical

Instead of guessing technical terms, describe physical phenomena:

Sensory Description	AI Understands
"Like physical paper folding"	3D flip animation with perspective
"Glass-like, see-through"	Translucent overlay with blur
"Fat at ends, thin in middle"	Bezier curve with pinched center ("waist" shape)
"Jagged edges"	Anti-aliasing issue

Principle 6: Screenshot + Text = Maximum Clarity

Screenshots solve 90% of visual misunderstandings:

•Capture current state
•Annotate with arrows/circles pointing to problem areas
•Add brief text describing expected vs actual

This is especially critical for:

•UI alignment issues
•Color/contrast problems
•Animation timing
•Layout spacing

Principle 7: Iterative Refinement

Pattern for Complex Features

•Describe the goal (what it should look like/do)
•Review first implementation
•Provide specific feedback using the Golden Formula
•Repeat until satisfied

When Stuck

If AI keeps misunderstanding:

•Try different sensory descriptions
•Provide a reference image (screenshot from another app, design mockup)
•Break the request into smaller pieces
•State what you DON'T want

Anti-Patterns to Avoid

Anti-Pattern	Why It Fails	Better Approach
"Make it better"	No actionable criteria	"Increase contrast between X and Y"
"Like the other one"	Ambiguous reference	"Like the Classic widget's shadow effect"
"Fix the bug"	No reproduction steps	"When I rotate after 10s idle, black flash appears"
Technical guessing	Often wrong terminology	Describe what you see/want

Communication Checklist

Before sending a request, verify:

• Location/file specified
• Current vs expected behavior described
• Constraints mentioned (API, performance, dependencies)
• Screenshot attached (if visual issue)
• Sensory description used (if complex visual effect)