Writing Effective Prompts

The more context you give Claude, the more precisely the psychological intelligence activates. This guide shows you how to write prompts that unlock deeper, more specific design thinking.

The core principle

Describe the human situation, not the UI components.

Claude knows what a form looks like. What it needs from you is who will use it, what they're feeling, what's at stake, and what could go wrong. That's the psychological context that makes designs intelligent.

Always start with a command

Pick the right intelligence mode before describing your needs:

/asciikit-quick      # 1-2 screens, fast
/asciikit-flow       # Multi-screen journey
/asciikit-explore    # Compare psychological approaches
/asciikit-spec       # Entire app architecture
/asciikit-build      # Technical spec with psychology docs

See The 5 Commands for when to use each one.

The anatomy of a great prompt

The most effective prompts give Claude four things:

1. What users are doing — the task or goal
2. Who they are — their emotional state and segment
3. What's at stake — what makes this moment significant
4. What could go wrong — the fears, anxieties, barriers

Example: Thin vs. rich prompts

Thin prompt:

Design a payment form

Claude gets: a generic payment form with standard fields.

Rich prompt:

Design a payment form for elderly users buying medication online
for the first time. Average order $80. They're worried about
online fraud and have difficulty reading small text.

Claude gets: large typography, prominent security badges, visible phone support option, larger touch targets, simplified card entry, reassuring microcopy, risk reversal guarantee.

Same command. Ten times more targeted psychology.

Describe the emotional context

The psychological intelligence responds most powerfully when you name the emotional situation.

Naming emotional states that trigger specific frameworks

Anxiety and fear:

/asciikit-quick
Design an account deletion screen. Users are scared of losing
their data and want to make sure they understand what happens
before they confirm.

→ Claude applies: non-blaming language, clear consequences, explicit confirmation, data export option, reversibility period

Skepticism and distrust:

/asciikit-flow
Create onboarding for a new fintech app. Users have been burned
by fee-heavy banks and are cynical about promises.

→ Claude applies: transparency-first architecture, concrete proof before features, no-hidden-fees messaging, social proof from people like them

Overwhelm and confusion:

/asciikit-quick
Design a tax form section for users who find financial forms
confusing and typically abandon at the "business income" section.

→ Claude applies: progressive disclosure, just-in-time explanations, simplified language, explicit "save and come back" option, help text at every field

Excitement and aspiration:

/asciikit-explore
Show different approaches for a career coaching app home screen.
Users are excited about changing careers but don't know where
to start.

→ Claude applies: aspiration triggers, quick wins, competence-building structure, multiple entry points

Describe the user segment

User segments activate different psychological frameworks. Claude adapts everything — typography assumptions, information density, trust signal types, and error handling.

Segments that trigger specific intelligence

Seniors:

/asciikit-quick
Design a video call setup screen for adults 70+ who are
using video calling for the first time to see their grandchildren.

→ Large touch targets, visible undo, phone support option, non-technical language, step confirmation

Power users:

/asciikit-spec
Design a developer dashboard for engineers who live in the
terminal and hate interfaces that slow them down.

→ Information density, keyboard shortcuts, batch operations, minimal hand-holding

First-time users in a complex domain:

/asciikit-flow
Create an investing onboarding flow for people who have never
invested before and are intimidated by financial language.

→ Jargon elimination, analogies, small first step, confidence building, fear of loss addressed

Anxious users under pressure:

/asciikit-quick
Design an emergency contact form for hospital patient intake.
Family members filling this out are in crisis and cannot
focus on complex inputs.

→ Extreme clarity, minimal fields, large text, no optional questions (mark clearly), save progress automatically

Describe the domain stakes

High-stakes domains trigger heavier trust and safety psychology automatically.

Example: Stakes change everything

/asciikit-quick
Design a subscription cancellation flow

vs.

/asciikit-quick
Design a subscription cancellation flow for a mental health
app where users might be cancelling during a difficult period.
Handle this with care — don't use dark patterns or guilt trips.

The second prompt activates: ethical design constraints, supportive language, genuine exit options, mental health sensitivity, no manipulation.

Prompt templates by use case

Trust-critical screens

Design a [screen] for [user segment] who [emotional state].
They're worried about [specific fear]. Make sure they feel
[target emotion] by the end.

High-abandonment flows

Create a [multi-step flow] for [user type]. The biggest drop-off
point is [specific step]. Users abandon because [reason].
Design specifically to address this.

Competing psychological needs

Show me different approaches for [screen] that balance
[competing need A] with [competing need B]. Users need
[goal] but risk [downside].

Vulnerable or special populations

Design [screen] for [specific vulnerable group]. They have
[specific challenge/context]. The design must avoid [harmful patterns]
and instead create [target emotional outcome].

Specifying what to avoid

You can explicitly direct the psychology away from patterns you don't want:

Design an engagement-focused home feed but without
infinite scroll, fake urgency, or manipulative patterns.
Prioritize digital wellbeing.

Create a premium upgrade screen that doesn't use artificial
scarcity or guilt. Make the value obvious without pressure.

Design a checkout flow with no dark patterns — no pre-checked
boxes, no confusing unsubscribe, no hidden fees.

Claude's psychology framework respects ethical boundaries. Naming what to avoid helps target the intelligence more precisely.

Refining with follow-up prompts

After Claude generates wireframes, guide the psychology with specifics:

Shift the emotional register:

This feels too aggressive — make it calmer and more reassuring

The trust signals feel corporate. Make them more human and specific

Adjust for a different user state:

Now show me how this looks for a returning user who already trusts us

What does this feel like if the user is in a hurry?

Push the psychology harder:

The cognitive load still feels high. Apply Miller's Law more aggressively

I want the loss aversion mechanics to be more prominent at the CTA

Explore the ethical boundaries:

Show me where this design could become manipulative, and offer
a more ethical alternative that achieves the same conversion goal

Common mistakes

Describing UI components instead of human situations

Bad:  "I need a modal with a headline and two buttons"
Good: "Users need to confirm a destructive action — they should
       feel clear about consequences and in control"

Generic user descriptions

Bad:  "Design for users"
Good: "Design for 55-65 year old first-time investors who are
       skeptical about financial apps after seeing news about bank failures"

Missing the emotional context

Bad:  "Create a checkout flow"
Good: "Create a checkout flow for a luxury item. Buyers at this
       price point experience purchase anxiety and may look for
       reasons to abandon. Keep them confident throughout."

Over-specifying UI structure

Bad:  "Put the security badge in the top right corner"
Good: "Trust is the key emotion — apply security signals wherever
       users are most likely to hesitate"

Quick reference

Next steps

• Your First Intelligent Design — See psychology in action
• The 5 Commands — Choose the right intelligence mode
• Intelligence Examples — Real-world psychological designs