design:validate

design:validate — Validate Understanding

You are validating that the design problem is well-understood, stakeholders are aligned, context is fresh, and gaps are addressed before proceeding.

When to Use

Pre-agent: After /design:feed, before /design:run Post-research: After Wave 1 agents complete, before Wave 2

Determine Context

1. Read .design/DESIGN-STATE.md to determine current state
2. If "Context feeding" is complete but "Agents run" is pending → Pre-agent validation
3. If Wave 1 agents completed but Wave 2 pending → Post-research validation

If neither state matches:

"Nothing to validate right now. Run /design:feed to add context, then validate. Or run /design:run to complete Wave 1, then validate."

Pre-Agent Validation

1. Read Context

Read .design/BRIEF.md to understand all context sections added.

2. Stakeholder Mapping

Build stakeholder map from "Source:" labels:

MoM - Product Sync 2026-03-10 → Product Manager
Email from Engineering → Engineer  
User interview with Sarah → Customer
Slack thread - Design Critique → Designer

Update BRIEF.md by appending:

## Stakeholders
| Source | Role | Key Input |
|--------|------|-----------|
| [source] | [role] | [1-line summary] |

3. Conflict Detection

Scan context sections for contradictory statements:

• "focus on speed" vs "focus on stability"
• "simplicity" vs "more features"
• "users want X" vs "users don't like X"

If conflicts found, add to validation report:

### Potential Conflicts
- [conflict description]
- Suggestion: Discuss with team before proceeding

4. Context Age Analysis

Parse timestamps from context sections:

• < 7 days: Fresh ✅
• 7-30 days: ⚠️ Consider refreshing

• 30 days: ⚠️ Potentially stale

Add to validation report:

### Context Age
- [date] ([age]) - [status]

5. Understanding Confidence

Rate confidence by area:

Area	Confidence	Notes
Problem statement	HIGH/MEDIUM/LOW	Based on source agreement
User needs	HIGH/MEDIUM/LOW	Based on research depth
Technical constraints	HIGH/MEDIUM/LOW	Based on engineering input
Business goals	HIGH/MEDIUM/LOW	Based on strategy docs

6. Problem Reflection Prompt

Prompt designer:

## Problem Reflection

Before we run agents, confirm your understanding:

**What problem are we solving?**
[Synthesize from BRIEF.md]

**Who experiences this?**
[Users identified in context]

**If we solve this, what changes?**
[Designer answers]

**If we DON'T solve this, what happens?**
[Designer answers]

---
Ready to proceed? (yes/no/revise)

Save designer's responses to .design/PROBLEM-REFLECTION.md.

7. Sign-off Tracking

Initialize sign-offs in DESIGN-STATE.md:

## Sign-offs
- [ ] Product Manager - pending
- [ ] Engineering - pending  
- [ ] Designer - pending

Post-Research Validation

1. Read Research Output

Read .design/research/RESEARCH.md to get the synthesized problem statement.

2. Compare with Reflection

Read .design/PROBLEM-REFLECTION.md (from pre-agent validation).

Compare:

• Does research match designer's understanding?
• Are there significant deviations?

3. Validation Prompt

## Research Validation

Research has synthesized:

**[Problem Statement from RESEARCH.md]**

**Your earlier reflection:**
[From PROBLEM-REFLECTION.md]

**Quick Check:**
- [ ] Research matches my understanding
- [ ] This problem is worth solving (impact > effort)
- [ ] We have enough evidence to proceed

---
If all pass → proceed to Wave 2
If any fail → /design:feed more context

4. Confidence Update

Update confidence based on research findings:

## Understanding Confidence (Post-Research)

| Area | Confidence | Evidence |
|------|------------|----------|
| Problem statement | HIGH | Research synthesis confirms |
| User needs | MEDIUM | Research adds detail |
| [etc] | [rating] | [source] |

5. Critical Gaps

From research, surface unknowns that could change the solution:

### Critical Unknowns
- What happens if export fails? (edge case)
- How often is this used? (prioritization)
- Technical capacity? (scope)

Generate Validation Report

Create .design/VALIDATION-REPORT.md:

# Validation Report: [PROJECT]
Generated: [DATE]

## Status
- [x] Pre-agent validation: [date]
- [ ] Post-research validation: pending

## Stakeholders
[Stakeholder table]

## Conflicts
[Any conflicts detected]

## Context Health
[Age analysis]

## Confidence Score
[Confidence table]

## Critical Gaps
[Unknowns]

## Sign-offs
[Who approved]

## Next Step
[Proceed / Need more context / Revise]

Mark Validation Complete

Update .design/DESIGN-STATE.md:

• If pre-agent: mark "Pre-agent validation: complete"
• If post-research: mark "Post-research validation: complete"

Output

• Updated .design/BRIEF.md (stakeholder section)
• Updated .design/GAPS.md (confidence scoring)
• Updated .design/DESIGN-STATE.md (validation status)
• New .design/PROBLEM-REFLECTION.md (designer's answers)
• New .design/VALIDATION-REPORT.md (full report)

Usage

/design:validate

Platform

Claude Code, OpenCode CLI