Generative Media Skills
Multimodal image, video, and audio generation skills for AI agents like Claude Code and Cursor, powered by muapi.ai.
Evidence shows: repository claims not to store credentials, API key passed via environment variable, and SECURITY.md states no direct service connection. However, no user confirmation mechanism, limited data flow transparency (only mentions uploading local files to CDN), insufficient sensitive data handling details. Dependency security only via GitHub Action scan, no specific dependency list or vulnerability report. External effects include media generation and file upload, but no rollback mechanism. Source attribution only via related project links in README, no clear third-party code provenance. Thus, most criteria scored low.
Evidence shows: README and SECURITY.md consistent, architecture description clear, but no error handling or failure message examples. Dependency availability relies on external muapi-cli and muapi.ai service, no offline or fallback plan. Thus, self-consistency scored higher, but dependency availability and failure messages scored lower.
Evidence shows: explicitly targets Claude Code, Cursor, Gemini CLI, etc., provides recipes and expert library for various scenarios, capability boundaries via model list and parameter validation. But trigger precision not clearly specified, environment fit only mentions compatibility, no detailed configuration. Thus, most criteria scored medium.
Evidence shows: clear information architecture, detailed install notes, rich examples, clear license. But naming stability not specified, known limitations not explicitly listed, versioning/changelog missing, maintenance responsibility only partially via SECURITY.md. Thus, most criteria scored medium, but versioning and changelog scored 0.
Evidence shows: output usability via --download and --view, marginal value via rich recipes and expert library, but cost-benefit not provided with pricing or cost comparison. Thus, output usability and marginal value scored higher, cost-benefit scored lower.
Evidence shows: claims in README not independently verified, related project links provide partial corroboration, but facts and inferences not clearly separated. Thus, all criteria scored low.
- This repository depends on external muapi-cli and muapi.ai service; functionality limited if service unavailable.
- No user confirmation mechanism; API key may be used automatically, configure with caution.
- No rollback mechanism; generation operations may be irreversible.
- No versioning/changelog; maintenance responsibility unclear.
What does this agent do, and when should you use it?
Generative-Media-Skills is a multimodal media generation skill library for AI agents, offering schema-driven CLI scripts and an MCP server to generate, edit, and display professional-grade images, videos, and audio. The repository is structured into core primitives (wrappers for file upload, image editing, and platform interactions) and an expert library with 41 end-to-end workflow recipes, ranging from 3D logo animations and UGC videos to UI designs and logo creation. All operations rely on the muapi-cli and MuAPI’s proprietary API, requiring an API key. It supports Claude Code, Cursor, Gemini CLI, and OpenCode, and includes an MCP server that exposes 19 tools for compatible agents.
The repository provides a set of CLI scripts and MCP server configurations to invoke MuAPI’s generation models from AI agents. It uses the muapi command-line tool to perform text-to-image, text-to-video, image-to-video, audio creation, background removal, face swap, lip sync, and clipping extraction. Workflows are defined in SKILL.md files that the agent reads and executes. Core scripts support --output-json and --jq for programmatic output handling. It supports local file uploads, automatic result download, and direct viewing.
- Developers using Claude Code or Cursor who want to generate product promo videos without writing API boilerplate
- Content creators needing quick social media images and videos, such as Instagram posts or YouTube thumbnails
- Marketing teams using the UGC Ads workflow to turn selfies and product photos into spoken video ads
- UI/UX designers using the UI Designer skill to quickly prototype mobile or web mockups
- E-commerce operators using the Amazon Product Listing pack to generate complete listing image sets
- Video editors using AI Clipping to automatically extract top-ranked short clips from long videos
What are this agent's strengths and limitations?
- Access to 100+ AI models including Midjourney v7, Flux Kontext, Seedance 2.0, Kling 3.0, and Veo3 for image, video, and audio generation
- Includes 41 ready-to-run workflow recipes such as 3D Logo Animation, UGC Video Factory, and AI Clipping for practical use cases
- CLI scripts output structured JSON with
--jqfiltering, making them easy to integrate into agentic pipelines - The MCP server exposes 19 tools, allowing MCP-compatible clients like Cursor to call generation capabilities without shell scripting
- Core dependency on MuAPI’s proprietary API and muapi-cli creates vendor lock-in; cannot migrate to other generation backends
- Requires a valid MuAPI API key and usage is metered, limiting free use
- Skill installation and execution require Node.js and npm, which may be an additional dependency for pure Python environments
- Documentation does not provide pricing details; cost estimation requires visiting the MuAPI dashboard
How do you install or deploy this agent?
- Install Node.js and npm (recommended) or Python.
- Install muapi-cli globally via npm:
npm install -g muapi-cli(orpip install muapi-cli). - Configure your MuAPI API key: run
muapi auth configureand enter your key, or directly usemuapi auth configure --api-key "YOUR_MUAPI_KEY". Get your key at https://muapi.ai/dashboard. - Install skills to your agent: run
npx skills add SamurAIGPT/Generative-Media-Skills --allto install all, or use--skillfor a specific one.
How do you use this agent?
After installation, you can use the skills via CLI or the MCP server. For example, generate an image: muapi image generate "a cyberpunk city at night" --model flux-dev. To run an expert skill like Nano-Banana, execute bash library/visual/nano-banana/scripts/generate-nano-art.sh --file ./image.jpg --subject "a glass hummingbird" --style "macro photography" --resolution "2k" --view. For MCP integration, run muapi mcp serve and configure an MCP client like Claude Desktop or Cursor to point to that server. You can also use piping: generate_prompt | muapi image generate - --model flux-dev.
How does this agent compare with similar options?
Compared to Open-Generative-AI, a free self-hosted AI media studio GUI offering similar model set without subscription, this repository provides more flexible CLI and agent integration.
FAQ
What credentials are required to use these skills?
muapi auth configure and obtain from https://muapi.ai/dashboard.Can all skills run offline?
How do I integrate the skills with Cursor?
muapi mcp serve and add the server to Cursor’s MCP configuration. Alternatively, use npx skills add with -a cursor to install the skills.Can the generated content be used commercially?
What happens if a model call fails?
--no-wait and poll for results, or check error outputs and exit codes. The skill scripts include semantic exit codes to facilitate agent handling.