Overview

Create dt-method-03-deep.instructions.md, the on-demand deep instruction file for Method 3: Input Synthesis. Provides advanced techniques for synthesizing research data into patterns, insights, and actionable problem statements when the standard method-tier instruction is insufficient.

Target File

.github/instructions/dt-method-03-deep.instructions.md

Frontmatter

---
description: 'Deep expertise for Method 3: Input Synthesis — advanced affinity analysis, insight frameworks, and problem statement articulation'
applyTo: ''
---

Required Content

Advanced Affinity Analysis

• Multi-pass clustering: First pass by theme, second pass by stakeholder, third pass by severity, reveals different insights each time
• Cross-stakeholder pattern detection: Identifying themes that span multiple stakeholder groups (shared pain points across departments)
• Outlier investigation: Data points that don't fit clusters are often the most interesting, investigate before discarding
• Temporal pattern recognition: Patterns that emerge across time dimensions (issues that appear at shift changes, seasonal patterns, onboarding-related vs. ongoing)

Insight Framework

Moving from observations to insights to actionable HMW questions:

• Observation → Inference → Insight formula: "[User group] [behavior observed] because [underlying need/motivation], which means [implication for design]"
• Insight quality criteria: An insight must be surprising (non-obvious), generative (opens design directions), and evidenced (traceable to research data)
• Insight vs. observation test: Coach helps user distinguish "operators skip step 3" (observation) from "operators skip step 3 because the feedback delay makes it feel unnecessary, which means the system's response time shapes compliance behavior" (insight)

HMW Question Scaffolding

• Breadth vs. depth calibration: HMW questions that are too broad ("How might we improve everything?") or too narrow ("How might we change button color?") need recalibration
• Generative tension: Good HMW questions contain a creative tension, "How might we make safety compliance feel empowering rather than bureaucratic?"
• HMW family generation: From one insight, generate 5-8 HMW questions varying the angle, stakeholder, constraint, aspiration, inversion
• Priority weighting: Lightweight technique for identifying which HMW questions have the highest design potential

Problem Statement Articulation

• Point of View (POV) format: "[User] needs [need] because [insight]", structured but not rigid
• Scope validation: Does the problem statement match the scope boundaries from Method 1?
• Assumption audit: Which assumptions in the problem statement are validated vs. still assumed?
• Multiple POV technique: Write problem statements from different stakeholder perspectives, identifies alignment and conflicts

Manufacturing Synthesis Patterns

• Process-vs-people clustering: Manufacturing findings often split into process improvement and human factors, synthesize across both
• Safety insight extraction: Special handling for safety-related findings (never deprioritize, always flag)
• Efficiency paradox detection: When "efficiency improvements" create downstream problems or shift burden to other stakeholders

Token Budget

Target: ~2,000-3,000 tokens (on-demand tier)

Source Material

Attach these files as context for the task-researcher phase:

• DT4HVE Method guidance: design-thinking-for-hve-capabilities/guidance/03-input-synthesis.md, canonical Method 3 with manufacturing synthesis patterns (L59)
• Coaching hats: design-thinking-for-hve-capabilities/.github/chatmodes/hats/, method-specific coaching hat files
• Cumulative research: Design Thinking cumulative research, Parts 1-2 (methodology foundations and cross-industry evidence) and Part 7 (cross-industry strategy and manufacturing reference patterns)

RPI Pipeline Workflow

1. task-researcher: Gather DT4HVE advanced Method 3 content, synthesis frameworks, insight formulation patterns, HMW scaffolding, manufacturing synthesis examples.
2. task-planner: Plan the file, affinity techniques, insight framework, HMW generation, problem statement formats, industry patterns.
3. task-implementor: Author following prompt-builder standards. Synthesis guidance should help users think more deeply, not automate the thinking.
4. task-reviewer: Validate synthesis methodology depth, insight quality criteria, HMW scaffolding effectiveness, prompt-builder compliance.

Starter Prompts

Research: /task-research — "Gather advanced Method 3: Input Synthesis content from DT4HVE — affinity analysis techniques, insight formulation, HMW question scaffolding, manufacturing synthesis patterns. Attach guidance/03-input-synthesis.md and cumulative research Parts 1-2, 7 as context."

Plan: /task-plan — "Plan dt-method-03-deep.instructions.md — multi-pass affinity analysis, insight framework with observation-inference-insight formula, HMW generation, problem statement formats."

Implement: /task-implement — "Author dt-method-03-deep.instructions.md following prompt-builder standards with empty applyTo, guidance-over-commands tone, 2000-3000 token budget."

Review: /task-review — "Validate dt-method-03-deep.instructions.md against prompt-builder standards, coaching tone, and synthesis methodology depth."

Prompt-Builder Compliance Checklist

Per .github/instructions/prompt-builder.instructions.md:

• description: frontmatter present and descriptive
• applyTo: '' (empty, on-demand loading)
• Writing style uses guidance over commands (no "You must/will/shall")
• Token count within budget tier
• No ALL CAPS emphasis
• Content structured for coaching context

Success Criteria

• File created at .github/instructions/dt-method-03-deep.instructions.md
• Frontmatter has empty applyTo: (on-demand loading)
• Advanced affinity analysis (multi-pass, cross-stakeholder, outlier, temporal)
• Insight framework with observation → inference → insight formula
• HMW question scaffolding with breadth/depth calibration
• Problem statement articulation with POV format and scope validation
• Manufacturing-specific synthesis patterns
• Token count within ~2,000-3,000 target
• Passes task-reviewer validation against prompt-builder standards
• Artifact registered in collections/design-thinking.collection.yml with path and kind fields
• Artifact registered in collections/hve-core-all.collection.yml with path and kind fields
• npm run plugin:generate run after updating collection manifests