abstract-to-concrete-design

Abstract to Concrete Design

An agentic CLI workflow that moves designers from scattered context to evidence-based design brief — before they open Figma.

---

The Problem: Scattered Context

Designers face a common challenge:

• MoMs in Slack threads
• Research in Notion
• Stakeholder feedback in email
• Competitive intel in random docs
• Design critiques in head

Context lives everywhere. The designer holds it all in their head — or doesn't. Important insights get lost. Decisions get made without full picture.

This system brings context together, validates understanding, and surfaces gaps — before design begins.

---

How It Works

/design:new         → Create project workspace
       ↓
/design:feed        → Paste MoMs, emails, docs, Slack threads
       ↓
/design:validate    → Verify understanding, detect conflicts
       ↓
/design:run         → Run agents (Research → Competitive → Critique)
       ↓
/design:validate    → Validate research synthesis
       ↓
/design:brief       → Get your evidence-based brief
       ↓
Open Figma

---

The Scattered Context Solution

Before	After
MoMs in Slack	All context in .design/BRIEF.md
Research in Notion	Synthesized in RESEARCH.md
Competitive intel scattered	Mapped in COMPETITIVE.md
Feedback in emails	Traced to source in BRIEF.md
"What do we know?"	Check GAPS.md or run /design:status

---

Key Features

1. Context Centralization

Paste anything — MoMs, emails, research docs, Slack threads. The system accumulates all context in one place, traceable to source.

2. Validation Gates

Before agents run, validate your understanding:

• Stakeholder mapping: Who said what?
• Conflict detection: Do stakeholders contradict each other?
• Context freshness: How old is this information?

3. Gap Detection

The system surfaces what's missing:

• No user research data?
• No engineering constraints?
• Missing competitive context?

You'll know what you don't know.

4. Learning Loop

The system remembers past issues:

• "URL was unreachable last time"
• "This context contradicted earlier input"

Future runs learn from history.

---

Commands

Command	Purpose
/design:new	Create project workspace
/design:feed	Add context (MoM, email, docs, Slack)
/design:validate	Verify understanding, detect conflicts, check freshness
/design:run	Execute research agents
/design:brief	Generate design brief
/design:status	Check current state and next step

---

Workspace Structure

.design/
  BRIEF.md              ← All context, traceable to source
  GAPS.md               ← What we know vs missing
  DESIGN-STATE.md       ← Workflow state tracker
  VALIDATION-REPORT.md  ← Problem validation results
  PROBLEM-REFLECTION.md ← Your confirmed understanding
  LEARNING-LOG.md       ← Past issues learned
  research/
    RESEARCH.md         ← Problem statement
    COMPETITIVE.md       ← Landscape analysis
    CRITIQUE.md          ← UX audit
    IDEATION.md          ← Design tensions
  DESIGN-BRIEF.md       ← Final output for Figma

---

Validation: Before Agents Run

/design:feed
> Paste MoM from product sync...

/design:validate
> ## Stakeholders
> | Source | Role | Key Input |
> |--------|------|-----------|
> | MoM - Product Sync | PM | Users can't find export |
>
> ## Conflicts
> None detected
>
> ## Context Age
> 2026-03-14 (today) ✅ Fresh
>
> ## Problem Reflection
> What problem are we solving? [Your answer]
>
> Ready to proceed? (yes)

---

Validation: After Research

/design:run (Wave 1 complete)
/design:validate
> ## Research Validation
> Research says: "Users struggle to locate export..."
>
> Your earlier reflection: "Users can't find export button"
>
> ✓ Matches my understanding
> ✓ Worth solving
> ✓ Enough evidence
>
> Proceed to Ideation? (yes)

---

The Agents

Research Agent

"What problem are we actually solving?" — Synthesizes context into a clear problem statement with user needs and constraints.

Competitive Agent

"What does the landscape look like?" — Maps competitors, table stakes, gaps, and opportunities. Descriptive, not prescriptive.

Critique Agent

"Where is the current experience failing?" — Audits your product via URL/screenshot. Severity-rated findings + what's working.

Ideation Agent

"What tensions must we navigate?" — Synthesizes into design tensions and provocations. Ends with questions, not answers.

---

Design Philosophy

• Agents surface evidence — not directions
• Competitive shows what exists — not what to copy
• Critique references specifics — not vague opinions
• Ideation opens thinking — doesn't close it

The designer brings the solutions. Agents provide the raw material.

---

Platforms

• Claude Code
• OpenCode CLI

---

Installation