Full AI-powered interior design consultation — room analysis, style recommendation, color palette, materials, and furniture suggestions. Use when the user wants design advice or a complete design brief.
Provide comprehensive design consultation by analyzing requirements and producing a complete design brief.
Before asking questions, check what data already exists to avoid redundant work.
!ls projects/${PROJECT_NAME}/brief.md 2>/dev/null && echo "HAS_BRIEF=true" || echo "HAS_BRIEF=false"
!ls projects/${PROJECT_NAME}/style-config.yaml 2>/dev/null && echo "HAS_STYLE_CONFIG=true" || echo "HAS_STYLE_CONFIG=false"
!ls projects/${PROJECT_NAME}/references/preprocessed/semantic_analysis.json 2>/dev/null && echo "HAS_SEMANTIC=true" || echo "HAS_SEMANTIC=false"
!ls projects/${PROJECT_NAME}/notes.md 2>/dev/null && echo "HAS_NOTES=true" || echo "HAS_NOTES=false"
!ls projects/${PROJECT_NAME}/references/*.{jpg,jpeg,png,webp} 2>/dev/null && echo "HAS_REFERENCE=true" || echo "HAS_REFERENCE=false"
HAS_SEMANTIC=true?
└── Read semantic_analysis.json — pre-fill style, colors, materials, lighting from Gemini Vision analysis.
Skip questions about things already known. Announce what was pre-filled.
HAS_BRIEF=true AND brief has content?
└── Read brief.md — pre-fill from existing brief. Only ask about missing/unclear sections.
HAS_STYLE_CONFIG=true?
└── Read style-config.yaml — style, palette, materials already decided. Focus on gaps.
HAS_NOTES=true?
└── Read notes.md — check for previous feedback to avoid recommending rejected options.
HAS_REFERENCE=true AND HAS_SEMANTIC=false?
└── Suggest running /preprocess-room first — Gemini Vision analysis will give much better data for consultation.
Never ask what the data already shows. Pre-fill and confirm instead of re-asking.
Ask the user (if not provided via $ARGUMENTS):
Based on requirements, recommend 2-3 suitable styles with reasoning:
Reference /style-guide for style characteristics.
Generate a cohesive color palette:
Recommend materials appropriate to the style:
Suggest key furniture pieces:
Recommend lighting layers:
Compile everything into a structured design brief:
## Design Brief: [Room Name]
### Style: [Primary Style] (with [Secondary] influences)
### Color Palette: [colors with hex codes]
### Key Materials: [top 5 materials]
### Furniture Plan: [key pieces]
### Lighting: [approach summary]
### Mood: [atmosphere description]
### Next Steps: [suggest using /generate-prompt or /mood-board]
After delivering the brief, suggest next actions:
/mood-board to generate a visual mood board description from this brief/generate-prompt to create AI rendering prompts based on this brief/render to send prompts directly to AI providers/mood-board — natural next step after consultation. Takes your design brief and expands it into a rich sensory description. Use when the client wants to "feel" the space before rendering./generate-prompt — skip mood board and go straight to prompts. Use when requirements are already detailed enough or the user is in a hurry./style-guide — reference during Step 2 (style recommendation). Essential when the user's style preference is vague or when recommending fusion styles./render — jump straight to rendering if the user already has a clear prompt. Rarely used directly after consult — usually goes through /generate-prompt first./edit-design — if the user arrives with a reference image instead of starting from scratch, redirect to /edit-design. Don't try to consult on an existing image here.