Skip to content
Agentmart
Submit a tool

image-generate

Make an image from a description with FLUX.2 klein on moria's GPU, about 20 seconds, and attach it.

Skill by GregHilston

IndexedMIT★ 15
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/greghilston-image-generate

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Make an image from a description with FLUX.2 klein on moria's GPU, about 20 seconds, and attach it. Use when asked to draw, paint, imagine, picture or generate an image or art, and for follow-ups that change one (make them zebras, wider, at night).

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/image-generate

What's inside

Skill name
image-generate
Description
Make an image from a description with FLUX.2 klein on moria's GPU, about 20 seconds, and attach it. Use when asked to draw, paint, imagine, picture or generate an image or art, and for follow-ups that change one (make them zebras, wider, at night).
License
MIT
Instructions
55 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Image generation

One command, already on `PATH`. Run it in your shell (the `terminal` tool in
Hermes, Bash in Claude Code, `bash` in pi) exactly as shown: no `python3` in
front and no directory. It is not a tool itself.

```bash
comfy-image.py 'PROMPT'                              # 1024x1024, a new random seed
comfy-image.py --seed 531322723 'PROMPT'             # same composition, changed prompt
comfy-image.py --width 1344 --height 768 'PROMPT'    # landscape; 768x1344 portrait
```

Put the prompt in single quotes, and write it without apostrophes ("the dog is",
not "the dog's"), so the shell passes it through whole.

## Write the prompt yourself

Never pass the user's words through as they are. Turn them into one paragraph
of plain description, 40 to 80 words: the subject and what it is doing, the
setting, the medium or style, the lighting, the colours, the composition.
Describe only what should be in the picture; there is no negative prompt.

> watercolor of buffalo playing basketball →
> A loose watercolor painting of two shaggy American bison playing basketball
> on an outdoor court at golden hour, one rearing up for a slam dunk while the
> other leaps to block, a chain-link fence and autumn trees behind, soft
> wet-on-wet washes, visible paper texture, warm ochre and blue palette

Words that must appear in the picture go in double quotes inside the prompt
(still inside the outer single quotes), and keep them short: the model
misspells long text.

## Follow-ups change the last image

When the user asks for a change ("make them zebras", "at night", "wider"),
take the prompt and seed of the last image from your own earlier reply, edit
only what they asked for, and run again with that `--seed` so the rest stays
put.
Use a new seed only when they want something different or another try.

## Reply

The output ends with a `MEDIA:` line. Start your reply with the prompt and
seed you used, as `Prompt: ... (seed 531322723)`, so the next change can reuse
them, then:

- **Hermes**: end your reply with that `MEDIA:` line exactly as printed, on its
  own line. That is what attaches the image.
- **Claude Code**: read the PNG with the Read tool to see it, and give the path.
- **pi**: give the path.

If it says moria is asleep or unreachable, say so; there is no other image
generator to fall back on. Do not use web search or `curl` for this.

From GregHilston/toolbox, MIT.

More from GregHilston/toolbox

Signals

Stars
15
Forks
5
Open issues
0
Last commit (this folder)
Oct 11, 2026 (17 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from GregHilston

Cite this page

image-generate skill. Agent Mart. https://agentmart.dev/skills/greghilston-image-generate (last updated 2026-10-24)