image-master — Claude Code skill for image prompts
The image-master skill structures prompts for realistic still images and checks them for common visual artifacts. Use it when generated faces, hands, or skin look distorted or artificial.
no license ★ 301 updated 2026-10-23
npx skills add hoodini/ai-agents-skills --skill image-master -a claude-codeWhat image-master does
image-master builds an image prompt from eight parts, covering the camera setup, subject, moment, setting, light, texture, composition, and mood. It pushes for concrete visual details instead of generic quality labels, then produces a self-contained prompt ready to paste into an image generator.
Before generation, the skill sets the viewpoint and reviews the prompt against known artifact risks. Reve, GPT-image, and Nano Banana mostly ignore negative prompts, so the request should describe the desired result directly. That matters: asking for “no extra fingers” can still draw attention to fingers. The skill focuses on still images; its description pairs nano-banana-2 with execution and director with motion.
What it helps with
- draft a prompt for a realistic still image
- describe natural skin or fur texture
- specify coherent light and shadows
- check an image prompt for anatomy artifacts
Example prompts for the agent
- Write a prompt for a realistic portrait with natural skin texture and side lighting.
- Build a wildlife photo prompt with a clear viewpoint, a decisive moment, and directional light.
- Review my image prompt for hand distortions and suggest specific changes.
How to install image-master
1Skills CLI (Vercel)
npx skills add hoodini/ai-agents-skills --skill image-master -a claude-code2Manually: put the files in place
~/.claude/skills/image-master/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/image-master/
1Skills CLI (Vercel)
npx skills add hoodini/ai-agents-skills --skill image-master -a codex2Manually: put the files in place
.agents/skills/image-master/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add hoodini/ai-agents-skills --skill image-master -a cursor2Manually: put the files in place
.cursor/skills/image-master/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install hoodini/ai-agents-skills skills/image-master2Manually: put the files in place
.github/skills/image-master/SKILL.md1Skills CLI (Vercel)
npx skills add hoodini/ai-agents-skills --skill image-master -a gemini-cli2Platform CLI
gemini skills install https://github.com/hoodini/ai-agents-skills --path skills/image-masterManually: put the files in place
.devin/skills/image-master/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about image-master
How do I install image-master in Claude Code?
Run npx skills add hoodini/ai-agents-skills --skill image-master -a claude-code. The skill is installed at ~/.claude/skills/image-master/SKILL.md.
What does image-master do?
It turns a scene idea into a detailed prompt for a still image and checks for common visual artifacts. The prompt follows eight sections, from camera setup to mood.
How does image-master handle negative prompts?
For Reve, GPT-image, and Nano Banana, it favors direct descriptions of the intended result because those tools mostly ignore negative prompts. Stable Diffusion, Flux, and Leonardo support them; Midjourney uses --no.