# Edit Images with Flux Kontext Pro

Precise image edits need clear model routing and focused prompts. This skill helps Claude, Codex, and Claude Code call Flux 1 Kontext Pro through RunComfy with practical prompt patterns.

## Install

```bash
npx skillstore add agentspace-so/flux-kontext
```

## Metadata

- Status: approved
- Slug: agentspace-so-flux-kontext
- Skillstore revision: r1
- Version status: missing
- Tree hash: 5f51b970ac251a9e590fc48a8390af7ea241e5801c81ef6a9a7733a87055706c
- Author: agentspace-so
- GitHub username: agentspace-so
- License: MIT
- Repository: https://github.com/agentspace-so/runcomfy-agent-skills/tree/main/flux-kontext/
- Ref: 1ffa7643651792ccb4bd3b15d924d2c97edff755
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/agentspace-so-flux-kontext
- Manifest: https://skillstore.pages.dev/api/skills/agentspace-so-flux-kontext/manifest

## Capabilities

- Documents the RunComfy endpoint for Flux 1 Kontext Pro image edits.
- Shows the required prompt and image input fields for a single-source edit.
- Provides example RunComfy CLI commands with output directory usage.
- Gives prompt patterns for preservation-led local image changes.
- Explains when to choose related models for batch, text, or generation tasks.
- Lists RunComfy CLI exit codes and troubleshooting references.

## Use Cases

- Refine a product image: Change a label, color, or small product detail while preserving lighting and composition.
- Create a controlled portrait edit: Add or adjust one visual element while keeping identity, pose, and clothing stable.
- Choose the right image model: Route single-image edits to Flux Kontext and redirect batch or text-heavy jobs to better models.

## Prompt Templates

### Basic object addition

```
Use Flux Kontext to edit this image. Keep the person, pose, and background unchanged. Add a small orange umbrella in the left hand.
```

### Preservation-led product edit

```
Use Flux Kontext for a product image edit. Keep the bottle shape, lighting, table, and camera angle unchanged. Replace only the label text with AURA.
```

### Reproducible variant run

```
Run Flux Kontext with a fixed seed. Keep the source identity and framing unchanged. Change only the jacket color to navy blue.
```

### Model-routing decision

```
Decide whether this image task should use Flux Kontext, Nano Banana Edit, GPT Image 2 edit, or Flux 2 Klein. Explain the choice, then prepare the best prompt.
```

## Limitations

- Supports one source image per edit workflow.
- Requires RunComfy CLI access and a RunComfy account or token.
- Uses public HTTPS image URLs for model input examples.
- Does not replace the latest RunComfy model page for new schema fields.

## Best Practices

- Lead with what must stay unchanged before describing the edit.
- Use one clear declarative edit instruction per model call.
- Set a seed when comparing repeatable variants.

## Anti Patterns

- Do not combine many unrelated edits in one prompt.
- Do not use this model for multi-image batch workflows.
- Do not send private images unless the user accepts RunComfy processing.

## Security Audit

- Audited at: 2026-07-09T04:05:52.155\+00:00
- Summary: The static findings are false positives from Markdown examples, hardcoded documentation links, placeholder URLs, and documented RunComfy CLI usage. I found no prompt injection attempt, covert exfiltration, unsafe shell interpolation, or malicious filesystem behavior in SKILL.md.

## Stats

- Views: 117
- Downloads: 11
- Favorites: 0
- Popularity score: 0