customer-centric-design
SKILL.md
Customer-Centric Design
This skill guides engineers and designers toward the right methodology for understanding users and validating solutions. It emphasizes solving real problems for real people.
Use-When
This skill activates when:
- Agent designs a new feature or product
- Agent asks "how do I know this is a real problem?"
- Agent needs to choose between qualitative vs quantitative research
- Agent wants to validate a design before building
- Agent asks "what metrics should I track?"
- Agent receives user feedback requesting a feature
- Agent needs to measure design impact
Core Rules
- ALWAYS validate the problem exists before designing a solution
- ALWAYS define success metrics before building
- ALWAYS choose methodology based on knowledge gap, not preference
- NEVER assume you know what users need without evidence
- ALWAYS validate design with actual users before shipping
Common Agent Mistakes
- Building features based on assumptions without problem validation
- Choosing research methods randomly instead of matching to situation
- Tracking vanity metrics that don't correlate with user success
- Skipping user validation because "it's obvious"
- Measuring activity instead of outcomes
Methodology Decision Tree
Use this decision framework to choose your approach:
Is the problem well-understood?
├── YES: Do users already complain about it?
│ ├── YES → Validate with analytics + quick usability test
│ └── NO → Optimize existing flow, measure impact
└── NO: Is this a new problem space?
├── YES → Qualitative research (interviews, observations)
└── UNCERTAIN → Hybrid: qualitative discovery + quantitative validation
Examples
✅ Correct
// Before: Validating the problem exists
// Step 1: Talk to 5 customers about the pain point
// Step 2: Check support tickets for related issues
// Step 3: Analyze product analytics for drop-off points
// Step 4: Define success metrics BEFORE building
const successMetrics = {
leading: "time-to-complete", // What we can move
lagging: "completion-rate", // The outcome
guardrail: "error-rate", // What we must not hurt
}
// Step 5: Validate with prototype before full build
// Run usability test with 5 users on Figma prototype
❌ Wrong
// Wrong: Building without problem validation
// "Users probably want dark mode, let's add it"
// Wrong: Choosing methodology arbitrarily
// "Let's run a survey" (when you need deep understanding)
// Wrong: Vanity metrics
analytics.track("button-clicks") // Activity, not outcome
// Wrong: No validation before shipping
// "It looks good, let's ship and see what happens"
References
- The Mom Test - Customer conversations
- Inspired: How to Create Products
- Google HEART Framework
- Lean Analytics
Weekly Installs
4
Repository
moderndegree/ag…t-skillsFirst Seen
Feb 17, 2026
Security Audits
Installed on
opencode4
gemini-cli4
github-copilot4
codex4
kimi-cli4
amp4