Agent Skills
Install MuAPI Agent Skills to generate AI images, videos, and audio directly from coding agents like Claude Code, Cursor, Gemini CLI, and other MCP-compatible tools.
Overview
MuAPI Agent Skills are pre-built, schema-driven shell scripts that give your AI coding agent the ability to generate, edit, and enhance rich media — all from the terminal.
- Pre-built skills for image, video, audio, effects, and more
- Dynamic Schema-Driven — scripts auto-resolve the latest models, endpoints, and valid parameters from
schema_data.json - Direct Media Display — use the
--viewflag to download and open generated media instantly - Local File Support — scripts automatically upload local images, videos, faces, and audio to the MuAPI CDN
- 100+ AI models — Midjourney v7, Flux Pro, Kling 3.0, Veo3, Suno V5, and more
Quick Start
1. Install Skills
# Install all MuAPI skills via the skills registry
npx skills add SamurAIGPT/Generative-Media-Skills --all
# Or install a specific skill
npx skills add SamurAIGPT/Generative-Media-Skills --skill muapi-media-generation
# List available skills
npx skills add SamurAIGPT/Generative-Media-Skills --list
2. Configure Your API Key
# Get your key at https://muapi.ai/dashboard
bash core/platform/setup.sh --add-key "YOUR_MUAPI_KEY"
3. Generate Your First Image
bash core/media/generate-image.sh \
--prompt "a cyberpunk cityscape at dusk, neon reflections on rain-slicked streets" \
--model flux-dev \
--view
4. Generate Your First Video
bash core/media/generate-video.sh \
--prompt "A timelapse of a city skyline transitioning from day to night" \
--model minimax-pro \
--duration 5 \
--view
Available Skills
MuAPI Skills are organized into three categories: Generation, Editing & Enhancement, and Platform utilities.
Generation Skills
| Skill | Script | Description |
|---|---|---|
| Text-to-Image | core/media/generate-image.sh | Generate images from text prompts. Supports Flux Dev, Midjourney v7, and 50+ models. |
| Text-to-Video | core/media/generate-video.sh | Generate videos from text prompts. Supports Minimax Pro, Kling, Veo3, and more. |
| Image-to-Video | core/media/image-to-video.sh | Animate a static image into a video. Supports Kling Pro, Veo3, Runway, and 15+ models. |
| Music & Audio | core/media/create-music.sh | Create music (Suno V5), sound effects (MMAudio), remix, and extend tracks. |
| File Upload | core/media/upload.sh | Upload local files (images, videos, audio) and get a CDN URL for use with other skills. |
Editing & Enhancement Skills
| Skill | Script | Description |
|---|---|---|
| Image Editing | core/edit/edit-image.sh | Prompt-based image editing with Flux Kontext, GPT-4o, Midjourney, and more. |
| Image Enhancement | core/edit/enhance-image.sh | One-click operations: upscale, background removal, face swap, colorize, Ghibli style, product shots, and more. |
| Lipsync | core/edit/lipsync.sh | Sync video lip movement to an audio track. Models: Sync Labs, LatentSync, Creatify, Veed. |
| Video Effects | core/edit/video-effects.sh | Apply effects to videos/images: Wan AI effects, face swap, dance animation, dress change, Luma modify/reframe. |
Workflow Skills (CLI Powered)
The AI Workflow ecosystem is now centrally powered by the muapi-cli which provides dedicated tools for LLM-based autonomous navigation.
| Skill | Command / Script | Description |
|---|---|---|
| Create & Edit | muapi workflow create<br>muapi workflow edit | Create or edit multi-node AI pipelines using natural language prompts. |
| Agent Discovery | muapi workflow discover | Fetch a full catalog of saved workflows (with AI-generated descriptions). Designed for LLMs to semantically match user intents to the best workflow. |
| Execution | muapi workflow execute | Execute a workflow directly via the API with custom inputs. |
| Interactive Run | muapi workflow run-interactive | Interactive CLI sequence that fetches input schemas and actively prompts the user for required values. |
Platform Skills
| Skill | Script | Description |
|---|---|---|
| Setup | core/platform/setup.sh | Configure API key, show config, and test key validity. |
| Check Result | core/platform/check-result.sh | Poll for async generation results by request ID. Supports one-shot and polling modes. |
How Skills Work
Each skill is a self-contained shell script powered by schema_data.json for dynamic model and endpoint resolution:
muapi-skills/
├── core/
│ ├── media/ # Generation primitives
│ │ ├── generate-image.sh
│ │ ├── generate-video.sh
│ │ ├── image-to-video.sh
│ │ ├── create-music.sh
│ │ └── upload.sh
│ ├── edit/ # Editing & enhancement
│ │ ├── edit-image.sh
│ │ ├── enhance-image.sh
│ │ ├── lipsync.sh
│ │ └── video-effects.sh
│ └── platform/ # Setup & polling
│ ├── setup.sh
│ └── check-result.sh
└── schema_data.json # Dynamic model registry
Key design principles:
- Scripts output clean JSON for seamless agentic integration
- All scripts support
--asyncfor fire-and-forget workflows - The
--viewflag auto-downloads and opens generated media schema_data.jsonvalidates models, resolves endpoints, and checks parameters at runtime
Local File Uploads
Most editing and media generation scripts now support a "File-First" workflow. Instead of providing a public URL, you can provide a local file path (e.g., --file ./photo.jpg or --face-file ./face.png).
The scripts will:
- Automatically upload the local file to the MuAPI CDN.
- Parse the resulting URL.
- Submit the generation/editing task using that URL.
- Clean up any temporary state (if applicable).
This ensures your AI agent can process media storage locally while leveraging cloud-based AI models.
Compatibility
Optimized for the next generation of AI development environments:
- Claude Code — Direct terminal execution via tools
- Gemini CLI / Cursor / Windsurf — Seamless integration as local scripts
- MCP — Each skill is Model Context Protocol-ready for universal agent usage
Common Flags
All core scripts share a consistent set of flags:
| Flag | Description |
|---|---|
--prompt, -p | Text description for generation |
--model, -m | Model name (script-specific defaults) |
--async | Return request_id immediately without waiting |
--view | Download and open the result (macOS) |
--file, --image-file, etc. | Local file paths for automatic upload (script-specific) |
--json | Raw JSON output only (for piping) |
--timeout N | Max wait in seconds before timing out |
--help, -h | Show usage and available options |
Related: Workflow Recipe Discovery
For agents that prefer natural-language recipes over shell-script primitives, the muapi assistant ships 40+ end-to-end workflow SKILL.md files (storyboard, action-figure-generator, ad-creative, etc.) and exposes them via a public, unauthenticated API:
GET https://api.muapi.ai/api/v1/agent-skills— list all recipesGET https://api.muapi.ai/api/v1/agent-skills/{name}— fetch full markdown body
See Agents API → Public Workflow Recipe Discovery for details.
Next Steps
- Get an API key at muapi.ai/dashboard
- Configure your agent —
bash core/platform/setup.sh --add-key "YOUR_KEY" - Start generating — Try the Quick Start examples above
- Explore models — Browse the Playground to discover 100+ available models