veryCoolTimo/imagegen-skills / skills/image-prompt
image-prompt
FlaggedUse when the user wants to turn a short idea into a rich, production-grade image-generation prompt — posters, landing-page or UI mockups, ads, editorial layouts, photoreal scenes, game screenshots, logos, or illustrations. Builds ONE structured, copy-paste-ready prompt optimized for OpenAI GPT Image 2.6 by default, following proven layered-composition patterns (named zones, hex palette, real font references, mood/finish, IP-safety). Triggers on requests like "сделай крутой промт для постера", "нужна картинка/постер/лендинг для…", "generate a prompt for a landing hero", "prompt for an ad / logo / game screenshot".
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 13.4 KB |
| references/anatomy.md | markdown | 7.5 KB |
| references/archetypes.md | markdown | 8.6 KB |
| references/edit-remix.md | markdown | 6.5 KB |
| references/fonts-palettes.md | markdown | 3.9 KB |
| references/generation.md | markdown | 3.1 KB |
| references/gold-examples.md | markdown | 23.2 KB |
| references/models/flux.md | markdown | 3.8 KB |
| references/models/gemini.md | markdown | 4.4 KB |
| references/models/gpt-image.md | markdown | 10.5 KB |
| references/models/ideogram.md | markdown | 3.3 KB |
| references/models/midjourney.md | markdown | 3.7 KB |
| references/models/recraft.md | markdown | 4.7 KB |
| references/models/reve.md | markdown | 3.9 KB |
| references/models/universal.md | markdown | 2.0 KB |
| references/styles.md | markdown | 5.2 KB |
references/models/flux.md
View on GitHub# Model adapter — FLUX (Black Forest Labs, FLUX.2 family)
How to render the 9-block content as a FLUX prompt. FLUX wants **flowing natural-language
prose** — flatten the content into a descriptive paragraph (JSON is only an internal scaffold,
see below). Output is prompt text only.
## Rules
- **Prose, not keywords.** Write like describing a scene to a person.
- **Specificity** drives quality — concrete attributes (age, material, texture, mood).
- **Composition order:** `[Subject] + [Action/Pose] + [Style/Medium] + [Setting] + [Lighting]
+ [Camera/Technical] + [Colors]`. A useful start, not a strict formula.
- **Front-load** the subject (and any quoted text) — earlier words weigh more.
- **Lighting is the highest-impact lever** — always specify (golden hour, Rembrandt,
volumetric fog, low-key chiaroscuro, studio softbox…).
- **NO negative prompts** — FLUX has none; negation makes it focus on the unwanted thing.
Replace "no X" with the positive alternative (no people → "deserted"; no text → "clean,
unmarked"; no blur → "tack-sharp").
- **Length:** ~30–80 words is the sweet spot (hard cap 512 tokens). English is most precise.
## Text & typography
- Quote exact text: `a poster with "HELLO WORLD" in bold letters`. Unquoted = description.
- Specify font class (sans/serif/script/display/mono), often with a reference ("bold geometric
sans-serif similar to Futura"); give size hierarchy + placement; front-load for accuracy.
- Keep strings short (1–4 words most reliable). Use **FLUX.2 [flex]** for best text.
## Hex color control (supported)
- **Always pair hex with a color name:** `#FF6B6B (coral pink)` — never bare hex.
- Bind each color to an object; **limit to 3–5 colors**; gradients ok
(`from #02EB3C to #EDFA3C`). Anti-bleed: `the LEFT chair in #…`, `maintaining exact color #…`.
## Model variants (FLUX.2)
`[klein]` fastest/high-volume (open, ≤4 refs, no prompt upsampling) · `[pro]` balanced default
(≤8 refs) · `[flex]` best for typography/posters/UI (exposes `steps` 1–50, `guidance` 1.5–10)
· `[max]` highest quality + real-time **grounding search** (≤8 refs). FLUX.1 legacy: Kontext =
editing, Fill = inpainting — use only when asked. **In FLUX.2 editing = T2I + reference images
(no separate edit model).**
## Editing & multi-reference
Pass source image(s) via `input_image`, `input_image_2`, … (URLs preferred). Phrase edits as
imperatives + explicit preserve list: `Change the background to a beach while keeping the
subject's pose, clothing, and expression exactly the same, maintaining the photo's realistic
style and color grading.` Multi-ref: assign each image a role by index ("face from image 1,
outfit from image 3, environment from image 5"). Break dramatic edits into sequential steps.
## Aspect ratios / resolution
Named: `1:1 16:9 9:16 4:3 3:2 21:9`; FLUX.2 also custom `width`/`height` (commonly ≤2048/side),
output up to ~4MP; editing input capped ~9MP total (pro API).
## JSON scaffold (optional)
For complex/multi-subject scenes, BFL suggests organising with a JSON object
(`scene / subjects[] / style / technical / colors`) — but **flatten it to flowing prose before
sending**. Keep JSON only as an internal or companion artifact.
## Output template (prose)
```
<subject + key attributes>, <action/pose>, <style/medium>, in <setting>.
<lighting>. <camera body + lens + aperture / film stock>. <composition>.
[palette: #HEX (name), #HEX (name)]. [quoted "TEXT" + font + placement].
```
**Report in assumptions:** `model=flux.2-<pro|flex|klein|max> · <WxH or aspect>`.
## Sources
docs.bfl.ml/guides/* and docs.bfl.ml/flux_2/*; official BFL Claude Code skill github.com/black-forest-labs/skills (`flux-best-practices`). Uncertain: exact per-model pixel bounds / whether enum vs raw width-height — defer to /api-reference.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |