Dt Rpi Integration

title: 'DT Image Prompt Generation' description: Guidance for writing lo-fi stick-figure visualization prompts during Method 5b Concept Articulation for stakeholder evaluation.

Purpose and Scope

During Method 5b (Concept Articulation), write visualization prompts that transform user concepts into lo-fi sketches for stakeholder evaluation.

Prompts are authored as part of YAML concept cards and optionally used with M365 Copilot or modern image models such as gpt-image-2 between Method 5b and 5c.

When to Offer Image Generation:

• After drafting concept title and description during Method 5b

• Before Method 5c Silent Review when visual concepts support stakeholder comprehension

• Only if lo-fi prompt structure is clearly explained and enforced

When Not to Offer:

• During Method 5a planning (concepts not yet articulated)

• If user prefers text-only concept evaluation

• If prompt violates lo-fi standards (redirect to structure before generating)

Prompt Generation Approach

Transform concept into visualization prompt using this sequence:

1. Identify Core Interaction: What single action or outcome does this concept test?

2. Extract Visual Elements: Who (stakeholder archetype by role), what (tool/object), where (environmental context from Methods 3-4)

3. Remove Personal Traits: Apply the representation safeguards below before writing the prompt

4. Constrain Text: Use only short, quoted labels when text is essential

5. Apply Lo-Fi Enforcement: Add all 5 required style directive layers

6. Validate 15-Second Test: Could this be sketched on a napkin in 15 seconds?

Concept → Prompt Workflow:

Concept: "Workers scan QR codes to access equipment manuals"
  ↓
Core Interaction: Worker scans code, sees instructions
  ↓
Visual Elements: Stick figure + phone + QR code + simple list
  ↓
Lo-Fi Layers: stick figures, minimal lines, plain background, b&w, no shading
  ↓
Prompt: "Create a simple stick-figure scene of a factory worker pointing
        a phone at a big QR code on a machine; on the phone screen, show
        a plain card with three bullet lines labelled 'Step 1, Step 2, Step 3'.
        Black-and-white line art, stick figures, minimal lines, no shading,
        plain white background."

Representation and Stereotype Safeguards

Concept prompts depict a role performing one interaction, never a specific person. Interview notes, personas, and customer cards often carry a participant's name, age, gender, ethnicity, or appearance; none of that belongs in a prompt by default.

• Describe the person by role and action: "an office manager reviewing a dashboard", "a technician scanning a machine".

• Leave out names, age, gender, ethnicity, body type, clothing, and other appearance details drawn from interviews or personas.

• Include a personal characteristic only when Observed or Reported evidence from Methods 2-3 shows it matters to the concept. Then describe the need or constraint rather than the trait: "a person holding a phone at arm's length to read small text", not "an elderly person".

• Do not infer a trait from a name, a role, or an age. An interview with a 50-year-old participant named Ashley becomes "a stick-figure office manager reviewing a dashboard", not "an older woman at a desk".

• Image models can fill in stereotyped defaults even for role-only prompts, such as drawing every nurse as a woman. After generation, review each image for stereotyped or one-sided depictions. Regenerate, changing one prompt constraint at a time, until the image reads as a neutral role sketch.

Image Prompt Structure

Required Template:

Create a simple stick-figure sketch showing [ONE CORE INTERACTION].
[Optional: Environmental context if constraint-relevant].
Minimal lines, plain white background, black-and-white line art, no shading.

Components:

• Opening: Action verb + "stick-figure" + subject (e.g., "Create a simple stick-figure sketch showing...")

• Single Scenario: One interaction per concept, not multiple use cases

• Optional Context: Environmental details only when relevant to constraints from Methods 3-4

• Terminal Reinforcement: Combine 3-5 lo-fi descriptors in closing sentence

• Text Constraint: Any on-image text must be short, quoted, and label-like; avoid sentences or dense UI copy

Subject, Context, Style, Focus, Exclusions:

• Subject: Stakeholder archetype by role (worker, nurse, manager) + tool/object, with no personal traits unless evidence requires them

• Context: Environmental constraints (factory floor, hospital bed, office desk) only when relevant

• Style: Always "stick-figure" or "simple line drawing" with "black-and-white line art"

• Focus: Single core interaction validating one key assumption

• Exclusions: "no shading", "no detail", "minimal lines" to block implementation detail

Lo-Fi Visual Requirements

5-Layer Required Directive Stack (all prompts must include all layers):

1. Human Representation: "stick figures" OR "simple line drawing"

2. Complexity Reduction: "minimal lines" (required)

3. Environmental Simplification: "plain white background"

4. Visual Treatment: "black-and-white line art"

5. Detail Blocking: "no shading" OR "no detail"

Effective Descriptor Vocabulary:

• Core: "stick figures", "minimal lines", "plain white background", "black-and-white line art", "no shading"

• Supplementary: "simple sketch", "napkin sketch", "whiteboard drawing", "basic shapes"

Anti-Pattern Vocabulary (never use):

• "detailed", "interface mockup", "buttons and menus", "production-ready", "polished", "realistic", "high-fidelity"

Scrappy Principle Alignment:

• Rough concepts invite structural feedback: "Does this solve the right problem?"

• Polished concepts invite cosmetic feedback: "I'd change the wording here"

• Maintain roughness to preserve validation quality

Method 5 Workflow Integration

Method 5b (Concept Articulation) — Prompts Written Here:

• Draft 2-4 word concept title

• Write 1-2 sentence description covering what and how

• Create visualization prompt with mandatory lo-fi structure

• Generate concepts.yml artifact with name/description/file/prompt fields

• Validate against 15-second napkin sketch test

Between Methods 5 and 6 — Images Generated (Optional):

• Optional generation: Use M365 Copilot or a modern image model to generate images from concepts.yml prompts

• Review generated PNGs and regenerate if outputs violate lo-fi standards or show stereotyped depictions of people, changing one prompt constraint at a time

Method 5c (Concept Evaluation) — Images Used for Alignment:

• Silent Review sequence with stakeholder representatives

• Visual concepts support 30-second comprehension test

• Three-lens D/F/V evaluation uses text + optional visuals

• All DT coaching artifacts are scoped to .copilot-tracking/dt/{project-slug}/. Never write DT artifacts directly under .copilot-tracking/dt/ without a project-slug directory.

YAML Concept Card Schema:

concepts:
  - name: Concept Title              # 2-4 words
    description: >-                   # 1-2 sentences, what + how
      Brief explanation of solution.
    file: concept-title.png           # kebab-case
    prompt: >-                        # Stick figures, minimal lo-fi
      Create a simple stick-figure sketch showing [core interaction].
      Minimal lines, plain white background, black-and-white line art, no shading.

Example Prompt Patterns

Example 1: Manufacturing Safety Concept

Concept: QR Snap Manuals — Workers scan machine QR codes to access step-by-step safety procedures.

Prompt:

Create a simple stick-figure scene of a factory worker pointing a phone at a big
QR code on a machine; on the phone screen, show a plain card with three bullet
lines labelled 'Step 1, Step 2, Step 3'. Black-and-white line art, stick figures,
minimal lines, no shading, plain white background.

Rationale: Single scenario (scanning QR code), environmental context (factory machine), all 5 lo-fi layers, passes 15-second sketch test.

Example 2: Healthcare Alert Concept

Concept: Vital Sign Pulse — Nurses receive automatic alerts when patient vitals exceed thresholds.

Prompt:

Create a simple stick-figure sketch showing a nurse holding a tablet with an upward
arrow and exclamation mark, standing next to a patient in a bed. Minimal lines,
plain white background, black-and-white line art, no shading.

Rationale: Healthcare environment (patient bed), single interaction (alert notification), simple symbols (arrow, exclamation), no UI mockup detail.

Example 3: Office Efficiency Concept

Concept: Dashboard Glance — Managers view team metrics on a single dashboard page.

Prompt:

Create a simple stick-figure sketch showing a person at a desk looking at a screen
with three simple bar charts. Minimal lines, plain white background, black-and-white
line art, no shading.

Rationale: Office context (desk), single interaction (viewing metrics), basic shapes (bar charts), avoids "detailed interface mockup" trap.