This skill provides a systematic framework for conducting brand identity research when a design feels "cheap," lacks character, or needs differentiation from competitors.
Inputs
User feedback on what feels "cheap" or generic
Current design URL or screenshots
3-5 reference URLs (same or adjacent industry)
Outputs
Research notes and artifacts under .agent/artifacts/{conversation-id}/
3+ strategic direction options with design DNA
Mockups (or documented fallback) per option
Comparison matrix and hybrid recommendation
Prerequisites
Ability to capture screenshots (manual or tool-based)
Stakeholder availability for decision-making
相關技能
Tools & Availability
browser_subagent for screenshots (if unavailable, capture manually)
generate_image for mockups (if unavailable, document prompt + create in Figma or skip with rationale)
When to Use
User reports current design feels generic or template-like
Design lacks distinct visual character
Need to evaluate if brand identity redesign is warranted
Want to develop strategic design direction options
Need to make data-driven design decisions
Methodology
Phase 1: Problem Definition
Capture User Feedback
What specific words does user use? ("cheap," "generic," "template")
What's the emotional gap? (aspirational vs actual perception)
What's the business impact? (pricing perception, conversion concerns)
Identify Key Questions
What makes designs feel "premium" vs "cheap"?
What creates "distinctive character"?
What design patterns do memorable brands use?
Is current approach (color, typography, layout) sufficient?
Phase 2: Reference Collection
Request References
Ask user for 3-5 landing pages they admire
Prioritize sites with "character" or "soul"
Mix of same-industry and aspirational cross-industry examples
Capture Screenshots
Use browser_subagent to navigate and capture
Get hero, features, testimonials, footer sections
Save with descriptive names: {sitename}_{section}_*.png
Document Current State
Capture existing design for before/after comparison
Focus on sections that feel weakest
Phase 3: Pattern Analysis
For each reference, analyze:
Color Palette
Base colors: Background and primary text
Accent colors: CTAs, highlights
Palette size: Single bold accent vs multi-color system
Emotional tone: Warm cream vs clinical white, dark vs light
Specific hex codes when visible
Typography
Font families: Serif vs sans-serif, display vs body
Visual weight: Thin/polite vs thick/bold
Scale: Moderate vs massive headlines
Unique elements: Handwriting, decorative fonts
Hierarchy patterns: How headings relate to body
Layout Patterns
Grid structure: Rigid vs broken/asymmetrical
Spacing: Tight vs generous whitespace
Visual volume: Thin borders vs thick, minimal vs maximal
Artifacts stored under .agent/artifacts/{conversation-id}/ with descriptive names
Each option includes color palette (hex codes), typography, signature, and risks
At least one mockup or a documented fallback per option
Comparison matrix + decision framework present
Pass/Fail: Pass only if all checks above are satisfied.
Risks & Mitigations
Reference copying risk → extract patterns, never replicate layouts verbatim
Overfitting to a single reference → require 3+ distinct references
Tool unavailability → document fallbacks and constraints in outputs
Common Pitfalls
❌ Don't: Develop 3 variations of the same safe direction
✅ Do: Create genuinely distinct options (safe, emotional, bold)
❌ Don't: Use vague descriptions like "modern" or "clean"
✅ Do: Specify exact hex codes, font names, element sizes
❌ Don't: Skip the hybrid recommendation
✅ Do: Most real projects need balanced combinations
❌ Don't: Generate mockups without detailed prompts
✅ Do: Include specific colors, fonts, moods in generate_image prompts
❌ Don't: Leave decision to gut feel
✅ Do: Provide star-rated comparison matrix and decision framework
Example Usage
User: "Our design feels cheap and generic."
Agent:
1. Request 3-5 reference landing pages they admire
2. Use browser_subagent to capture screenshots
3. Analyze each for color/typography/layout/texture/human patterns
4. Critique current design against reference standards
5. Develop 3 distinct options (Editorial, Human, Brutalist)
6. Generate mockups for each option
7. Create hybrid combining best of 2-3 options
8. Generate 2 hybrid variants (light/dark)
9. Create comparison matrix and decision framework
10. Document in comprehensive research notes
11. Request user selection to proceed with implementation