Generate AI image prompts with visual identity and style consistency — reads brand charter, applies artistic styles, outputs multiple prompt variants. Triggers: "image-prompt" | "generate image prompt" | "image prompt" | "prompt for image" | "visual prompt" | "generate a prompt for Grok" | "generate a prompt for Midjourney" | "generate a prompt for flux" | "image generation prompt" | "create image prompt" | "prompt for dall-e".
npx claudepluginhub roxabi/roxabi-plugins --plugin image-prompt-generatorThis skill is limited to using the following tools:
Let:
Guides creation of production-quality image prompts for Nano Banana AI generator through brand selection, concept exploration, crafting, and iterative revision. Activates on 'create image prompt' or blog image requests.
Enhances AI image generation prompts with Subject-Context-Style structure, style anchors, character consistency, and mcp-image workflows. Improves prompt quality for better results.
Generates production-ready GPT-Image2 prompts by selecting styles, templates, categories, and tags from the awesome-gpt-image-2 library. Use for creating, rewriting, or improving image-generation prompts.
Share bugs, ideas, or general feedback.
Let:
χ := ~/.roxabi-vault/config/visual-charter.json
φ := ~/.roxabi-vault/config/face-reference.json
FACE_DESC := face description (set in Phase 2.5 if user wants their face in image)
Goal: Transform a concept into multiple ready-to-use image generation prompts, optionally aligned to a brand visual charter.
charter="$HOME/.roxabi-vault/config/visual-charter.json"
if [ -f "$charter" ]; then
echo "CHARTER_FOUND: $charter"
cat "$charter"
else
echo "NO_CHARTER — using defaults (no brand constraints)"
fi
χ ∃ → parse brand identity: colors, style preferences, mood, avoidances.
χ ∄ → proceed without brand constraints — inform user they can create one from examples/visual-charter.example.json.
Check φ and load if ∃.
Concept ∄ via $ARGUMENTS → → DP(B) "What image do you want to create? Describe subject, context, any specific requirements."
Ask structured follow-up (one message, all at once):
Parse into creative brief: Subject | Context | Platform | Content type | Mood | Style | Aspect ratio
→ DP(A) "Do you want your face/likeness in the image?" Options: Yes | No
Execute only if user confirmed face in image.
mkdir -p "$HOME/.roxabi-vault/config"
cat > "$HOME/.roxabi-vault/config/face-reference.json" << EOF
{
"description": "FACE_DESCRIPTION",
"updated_at": "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
}
EOF
echo "=== Loading references ==="
plugin_dir=$(dirname "$(dirname "$(dirname "$0")")")
for ref in references/artistic_styles.md references/prompt_best_practices.md; do
if [ -f "$ref" ]; then
echo "LOADED: $ref"
fi
done
Read references/artistic_styles.md (available styles) and references/prompt_best_practices.md (prompt structure, platform tips).
Generate 4-6 variants from creative brief. ∀ variant includes:
| Component | Description |
|---|---|
| Style | Artistic style from references — user style preference → lead with that; ∄ → distribute across categories |
| Subject | Detailed description — FACE_DESC ∃ → prepend: "[FACE_DESC], [rest of subject]" |
| Composition | Framing/perspective matched to target aspect ratio |
| Lighting | Source and quality matched to mood |
| Color palette | Dominant colors/harmony — χ ∃ → informed by brand charter |
| Mood | Emotional tone matched to brief |
| Technical | Resolution, aspect ratio, platform-specific rendering |
Variant distribution: style preference given → 2-3 in that direction + 1-2 creative divergences. ∄ preference → 1-2 photographic, 1-2 illustration/digital art, 1 stylized, 1 experimental.
χ ∃ → apply to ∀ variant:
Mark each: On-brand / Near-brand / Off-brand (deliberate divergence).
Auto-save all variants before presenting:
save_dir="$HOME/.roxabi-vault/image-prompts"
mkdir -p "$save_dir"
date_prefix=$(date +%Y%m%d)
slug=$(echo "USER_CONCEPT" | tr '[:upper:]' '[:lower:]' | sed 's/[^a-z0-9]/-/g' | sed 's/-\{2,\}/-/g' | sed 's/^-\|-$//g' | cut -c1-40)
save_file="$save_dir/${date_prefix}_${slug}.md"
# Write concept, face reference (if used), and all variants to the file
echo "# Image Prompts — USER_CONCEPT" > "$save_file"
echo "Date: $date_prefix" >> "$save_file"
# Append each variant block
echo "Saved to: $save_file"
Present in structured format:
Variant 1 — [Style Name] [Brand Alignment]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Prompt: [full prompt text ready to copy]
Style: [style category]
Composition: [framing details]
Lighting: [lighting description]
Mood: [atmosphere]
Best for: [recommended platform/use case]
→ DP(A) Pick one (variant number) | Refine (variant + changes) | Regenerate (different styles) | Batch (run generate_prompt_variants.py)
python3 scripts/generate_prompt_variants.py \
--concept "user concept here" \
--charter "$HOME/.roxabi-vault/config/visual-charter.json" 2>/dev/null || \
python3 scripts/generate_prompt_variants.py \
--concept "user concept here"
Present batch results for selection.
$ARGUMENTS