← Back to list
🎨

flux-best-practices Checked

Comprehensive guide for BFL FLUX image generation models. Covers prompting, T2I, I2I, structured …
🎨 Design & Creative skills By calesthio Version v1.0.0 Added 2026-09-07 Updated 2026-09-07 Source ↗
8.2Overall rating
2Views
8Downloads
56.5kStars

Installation

🤖 Install via AI

Copy the prompt below and send it to your AI assistant (e.g. Claude Code) — it will follow the instructions and install automatically.

Install the "flux-best-practices" skill by following the instructions at https://skill123.me/install/flux-best-practices.
⌨️ Command line install

Run in your terminal — downloads and installs to ~/.claude/skills/.

curl -fsSL https://skill123.me/install/flux-best-practices.sh | bash
📦 Download ZIP

Download the zip and extract it into your skills directory (e.g. ~/.claude/skills/), then restart your session.

⬇ Download v1.0.0 · 50 KB
📊 Static rating
8.2/10
Trigger
8.0
Structure
8.7
Workflow
5.6
Content
9.0
Engineering
10.0
Security
10.0

Six-dimension rubric with reviewer rationale.

View scorecard ↓
🛡 Dynamic test
—

Not live-tested yet — this skill is rated on static analysis only.

About this skill

Comprehensive guide for BFL FLUX image generation models. Covers prompting, T2I, I2I, structured JSON, hex colors, typography, multi-reference editing, and model-specific best practices for FLUX.2 and FLUX.1 families.

FLUX Best Practices

Use this skill when generating prompts for any BFL FLUX model to ensure optimal image quality and accurate prompt interpretation.

> **Extended reference:** AGENTS.md (AGENTS.md) in this directory is the long-form upstream FLUX guide (vendored from Black Forest Labs). It is supplementary reference material scoped to this skill only — SKILL.md is the loadable entry point and the authority. It does not override or extend the repository-root AGENTS.md / AGENT_GUIDE.md.

Documentation

Comprehensive guide for BFL FLUX image generation models. Covers prompting, T2I, I2I, structured JSON, hex colors, typography, multi-reference editing, and model-specific best practices for FLUX.2 and FLUX.1 families.

FLUX Best Practices

Use this skill when generating prompts for any BFL FLUX model to ensure optimal image quality and accurate prompt interpretation.

> **Extended reference:** AGENTS.md (AGENTS.md) in this directory is the long-form upstream FLUX guide (vendored from Black Forest Labs). It is supplementary reference material scoped to this skill only — SKILL.md is the loadable entry point and the authority. It does not override or extend the repository-root AGENTS.md / AGENT_GUIDE.md.

When to Use

  • Creating prompts for FLUX.2 or FLUX.1 models
  • Text-to-image (T2I) generation
  • Image-to-image (I2I) editing with FLUX.2 models
  • Structured scene generation with JSON
  • Typography and text rendering
  • Multi-reference style transfer
  • Color-accurate brand generations

Quick Reference

### Prompt Structure Formula


[Subject] + [Action/Pose] + [Style/Medium] + [Context/Setting] + [Lighting] + [Camera/Technical]

### Model Selection

| Use Case | Recommended Model | Notes |

| ------------------- | ----------------- | -------------------------------------- |

| Fastest generation | FLUX.2 [klein] | 4B or 9B, sub-second |

| Highest quality | FLUX.2 [max] | Best detail, grounding search |

| Production balanced | FLUX.2 [pro] | Quality + speed |

| Typography/text | FLUX.2 [flex] | Best text rendering |

| Local/development | FLUX.2 [dev] | Open weights |

| Image editing | FLUX.2 [pro/max] | Pass image URL directly to input_image |

| Inpainting | FLUX.1 Fill | Object removal/completion |

| Context editing | FLUX.1 Kontext | Older model, prefer FLUX.2 |

### Critical Rules

1. **NO negative prompts** - FLUX does not support negative prompts; describe what you want

2. **Be specific** - Vague prompts produce mediocre results

3. **Use natural language** - Prose/narrative style works best

4. **Specify lighting** - Lighting has the biggest impact on quality

5. **Quote text** - Use "quoted text" for typography rendering

6. **Hex colors** - Use #RRGGBB format with color description

Related

For API integration (endpoints, polling, webhooks), see the **bfl-api** skill.

Rules Reference

Read individual rule files for detailed guidance:

View full documentation
  • rules/core-principles.md (rules/core-principles.md) - Universal FLUX prompting principles
  • rules/flux2-models.md (rules/flux2-models.md) - FLUX.2 family: klein, max, pro, flex, dev
  • rules/flux1-models.md (rules/flux1-models.md) - FLUX.1 family: older generation of FLUX.2 models - pro, Kontext, Fill
  • rules/t2i-prompting.md (rules/t2i-prompting.md) - Text-to-image prompting patterns
  • rules/i2i-prompting.md (rules/i2i-prompting.md) - Image-to-image editing with FLUX.2
  • rules/json-structured-prompting.md (rules/json-structured-prompting.md) - Complex scene composition
  • rules/hex-color-prompting.md (rules/hex-color-prompting.md) - Precise color specification
  • rules/typography-text.md (rules/typography-text.md) - Text rendering and typography
  • rules/multi-reference-editing.md (rules/multi-reference-editing.md) - Multi-image references
  • rules/negative-prompt-alternatives.md (rules/negative-prompt-alternatives.md) - Positive alternatives
  • rules/model-selection-guide.md (rules/model-selection-guide.md) - Choosing the right model

Example Prompt


A weathered fisherman in his 70s with deep wrinkles and a salt-and-pepper beard,
wearing a navy cable-knit sweater, standing at the helm of his wooden boat.
Golden hour sunlight from the left creates dramatic rim lighting on his profile.
Shot on Hasselblad with 85mm lens at f/2.8, shallow depth of field with harbor
lights creating soft bokeh in the background. Kodak Portra 400 color science.

Score breakdown

Trigger
8.0
Rich What/When and synonyms but no When-Not section, boundaries only implied by model scope.
Structure
8.7
84-line SKILL.md with 11 well-described rule files; rules/ naming plus vendored 2246-line AGENTS.md duplicates content.
Workflow
5.6
Formula and critical rules guide prompt construction, but no stepwise workflow, error handling or self-verification.
Content
9.0
Concise imperative writing, concrete prompt examples, consistent formatting throughout.
Engineering
10.0
Valid frontmatter, all referenced paths exist, no scripts to execute.
Security
10.0
No scripts, no network calls, no injection or secrets; only official bfl.ai URLs.
Clean, well-structured prompting knowledge base for FLUX models with strong progressive disclosure; lacks When-Not boundaries, execution steps and verification loop since it is purely a reference guide.