Scenario agent skills bring image, video, audio and 3D generation to coding agents
The scenario-labs/skills repository packages Agent Skills that route generation work through the Scenario MCP server, plus separate expert skills that drive Blender, Maya, ZBrush, Unreal and Unity.
What the repository is
scenario-labs/skills is a set of Agent Skills that, according to the README, teach AI coding agents “(Claude Code, Cursor, Codex, Copilot, and 70+ others)” how to create content with Scenario through the Scenario MCP server. The content types listed are images, video, audio, textures, skyboxes, 3D assets and custom-trained models. The skills follow the Agent Skills format.
The repository is MIT licensed, written mainly in Python, and was created on 2026-08-12. At the time of the GitHub trending snapshot it had 2958 stars, 674 of them gained in the last day, and 264 forks.
Two kinds of skills
Most of the repository is Scenario skills, which work through the Scenario MCP server. The README groups them by job. A getting-started skill, scenario, covers connecting to the MCP server and the core generation loop: discover, schema, run, wait, display, download. Other groups cover finding a direction before generating, images, game art and 3D, and video and audio. Examples from the tables include sprite sheets and walk cycles, tileable textures and PBR materials, 360 equirectangular skyboxes, text or image to 3D meshes with GLB/FBX download, comic pages and storyboards, and music, sound effects and speech.
The expert tools at the bottom of the list work differently. They are teams of skills that drive DCC (digital content creation) software and game engines installed on the local machine: ZBrush, Blender, Maya, Unreal Engine, Unity, Godot and Roblox Studio. The README says they are distilled from expert tutorials, conference talks and official documentation, need the application rather than the Scenario MCP server, and that each family folder’s README records how it was built and how far it was verified. A family installs as one set, since its specialists import the lead skill’s scripts.
Install and setup
Skills are installed with the skills CLI. To pick from a menu or install a single skill, the README shows:
npx skills add scenario-labs/skills
A single skill:
npx skills add scenario-labs/skills --skill scenario
The Scenario skills need the Scenario MCP server connected. For Claude Code the README gives:
claude mcp add --transport http scenario https://mcp.scenario.com/mcp
Other MCP clients can add the same URL and sign in with a Scenario account over OAuth, or use an API key. In practice, then, the Scenario skills depend on a Scenario account, while the expert tools depend on having the desktop application installed.
Pinning and updates
By default the install commands fetch the latest copy of each skill from main. To install a reviewed state, the README says to pin both the installer version and the commit taken from a release’s tag, and to run the command again with a newer commit to move.
Releases are staged by release-please. Merging the release PR tags a version, updates CHANGELOG.md, and publishes a GitHub release. To pull changes, the README says to run npx skills update or re-run the original install command. One limitation is noted: the menu’s topic groups cannot be selected from the command line, because --skill matches exact skill names only.
Source: scenario-labs/skills on GitHub