Image: GitHub
Why it mattersA design or game team can hand image, video and 3D jobs to a coding agent. The agent picks the right model and keeps a character the same across a batch.
A designer, game artist or marketing team can send image jobs to the same coding agent that writes code. Scenario released 80 MIT-licensed Agent Skills on GitHub. The skills teach Claude Code, Cursor, Codex and Copilot to work through its MCP server at mcp.scenario.com. Scenario says 70 other coding agents also work with the skills. The repository was ranked 25th on Trendshift's daily list on 3 November 2026.
What the skills do
Each skill is a short, versioned instruction set in the Agent Skills format. The README groups them by job. Finding a direction turns a brief into options and a moodboard. Images cover text-to-image, masked edits, product shots, brand kits, letter-perfect text overlays and storyboards. Game art covers sprites, tilesets, PBR textures, seamless skyboxes, image-to-3D meshes and motion capture from a reference video. Video covers text-to-video, color grading, music-video assembly and ad variants. Audio covers music, sound effects, voice and dubbing. The repository also carries skills for consistency across a set, custom model training, team admin and workflow authoring.
Expert tools run on your machine
Nine expert tool sets drive software a team already runs: Maxon ZBrush, Blender, Autodesk Maya, Unreal Engine, Unity, Godot and Roblox Studio. These skills do not call the Scenario service. They talk to the application through a local bridge. They read the scene, run operations and export files. Scenario says each bridge was proven live. The company adds that full tests of every skill are still pending.
How a team installs it
The install line in the README is one npx command that names every skill. INSTALL.md groups the same skills by role: 2D artist, 3D artist, producer, admin and others. Each role has one short command. Teams that want a reviewed state can pin a commit from a GitHub release tag. The latest release at the time of writing was skills-v0.60.1, published on 29 October 2026. The Scenario skills need a Scenario account. A team signs in with an API key or an OAuth sign-in to the MCP server.
The costs behind it
Scenario is a paid service. The company's own site calls Scenario "the leading AI models in one place for creative production and game assets". The skills themselves are MIT. The repository had 1,817 stars at the time of this item. Image, video and audio generation through Scenario is billed in the company's Creative Units. Nothing in the public repository states a per-unit price. A team that signs in pays Scenario for every generation the agent runs. The expert tool skills run locally and have no per-call fee from Scenario.
Why this matters
A team that already uses a coding agent gets a route to its own creative pipeline through the same window. The agent picks the right model for a shot. It prices the generation before spending. It keeps a character or product the same across a batch of images. The expert tool skills then let the same agent cut or render inside Blender, Maya or Unreal. The finished asset lands where the artist works. A team that wants to stay off a paid service can read every prompt. They can fork the repository and swap the Scenario calls for a local model.
Source
scenario-labs/skills, GitHub, MIT. Scenario MCP documentation.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
