scenario-labs/skills

Overview

Scenario Agent Skills is a curated collection of Agent Skills – small, reusable instructions that let AI coding assistants (Claude Code, Cursor, GitHub Copilot, etc.) interact with the Scenario MCP platform. Scenario provides a cloud service for generating production‑ready creative assets (images, video, audio, 3D models, textures, skyboxes, custom‑trained models) that are commonly used in games, marketing, and other entertainment pipelines.

The repository bundles 71 individual skills, each exposing a single capability (e.g., scenario-image, scenario‑video‑editing, scenario‑3d). By installing a skill with the npx skills add command, a developer equips an AI‑assistant with a new function it can call, passing parameters such as prompts, reference files, or style settings. The assistant then talks to the Scenario MCP server, runs the generation workflow, and returns the finished asset.

Key points

• Target audience – AI‑assisted creators, game‑dev teams, marketing producers, and anyone who wants an LLM to generate or edit visual/audio content automatically.
• Technology stack – Uses the Agent Skills specification (https://agentskills.io) and the Scenario MCP HTTP API. The skills are thin wrappers; the heavy lifting (diffusion models, video synthesis, 3D reconstruction, etc.) lives on Scenario’s backend.
• Installation model – Skills are installed via the skills CLI (npx skills add …). Users can install the whole suite, a role‑based subset, or individual skills.
• Extensibility – New skills can be added by following the same folder/SKILL.md pattern, making the collection grow as Scenario releases more asset types.

---

Installation & Usage

Prerequisites

1. Scenario MCP server – You must have access to a Scenario MCP endpoint (default https://mcp.scenario.com/mcp).
2. Authentication – Sign in with a Scenario account (OAuth) or supply an API key (found in the Scenario app settings).
3. skills CLI – The repository assumes you have the skills command line tool installed (provided by the Agent Skills ecosystem).

Quick start (install everything)

npx skills add scenario-labs/skills \
  --skill scenario \
  --skill scenario-image \
  ... (list all 71 skills) ...

This pulls the latest version of each skill from the main branch.

Role‑based installation examples

• Game art & 3D (22 skills) – installs sprite, texture, skybox, and 3‑D generation tools.
• Marketing & brand visuals (22 skills) – focuses on product shots, brand kits, ad variants, and thumbnail creation.
• Video & audio (23 skills) – gives text‑to‑video, video editing, music‑video assembly, and audio synthesis.
• ChatGPT & Codex pets – creates animated avatar assets for LLM‑driven chat agents.

Each role command is documented in INSTALL.md.

Connecting to the MCP

claude mcp add --transport http scenario https://mcp.scenario.com/mcp

or configure any MCP client with the endpoint and your API key. The scenario skill contains the full connection workflow.

Updating

Run npx skills update in a directory where you previously installed the skills, or repeat the original add command to fetch the newest definitions.

---

Core Concepts

Concept	What it means for users
Agent Skill	A tiny, self‑contained description (in SKILL.md) that tells an AI‑assistant how to call a specific Scenario API, what inputs it expects, and what output it returns.
MCP (Model‑Control‑Plane)	Scenario’s server that orchestrates generation jobs, stores assets, and provides status polling.
Lead skill	The scenario skill – establishes the basic discovery‑run‑wait‑download loop that all other skills build upon.
Sibling/lead‑skill relationship	More specialized skills (e.g., scenario‑video‑editing) invoke the core loop from scenario and add extra parameters or post‑processing steps.
Expert tools	Optional families that drive local DCC software (Blender, Unreal, Unity, etc.) – they are not part of the cloud MCP flow and require the corresponding application installed locally.

---

Notable Skills (selected)

Skill	Primary purpose
scenario‑image	Text‑to‑image generation and masked in‑painting; choose model, size, reference images.
scenario‑video	Text‑to‑video or image‑to‑video generation, lip‑sync, motion prompting, basic editing utilities.
scenario‑3d	Convert text or images into 3‑D meshes (GLB/FBX), supports multi‑view reconstruction and re‑texturing.
scenario‑game‑assets	Batch creation of sprites, icons, tilesets, and concept art with style consistency.
scenario‑audio	Generate music, sound effects, voice‑overs; includes utilities for scoring video.
scenario‑brand‑kit	Produces a full visual identity package (SVG logo, palette, type specs) from a brief.
scenario‑chatgpt‑pet‑create	Generates an animated “pet” avatar (sprite sheet, GIF, Codex package) from a concept or photo.

Each skill lives in its own folder with a SKILL.md that details the required parameters, optional flags, and example prompts.

---

Ecosystem & Integration

• AI coding assistants – The skills are designed to be called by Claude Code, Cursor, GitHub Copilot, etc., enabling a developer to write a prompt like “@scenario-image generate a cyberpunk cityscape” and have the assistant run the skill automatically.
• Scenario platform – All heavy AI models (Stable Diffusion, Runway, Gemini, ElevenLabs, etc.) are hosted by Scenario; the repository does not contain model weights.
• Agent Skills registry – The badge at the top links to the public skill index (skills.sh) where each skill is searchable and versioned.
• Release workflow – Changes are merged to main, then release‑please creates a GitHub release, updates CHANGELOG.md, and posts to the Scenario changelog page.

---

Strengths & Limitations

Strengths

• Comprehensive coverage – 71 skills span the whole creative pipeline (concept, asset generation, editing, final assembly).
• Modular – Users can pick only the skills they need, keeping the AI‑assistant prompt surface small.
• Up‑to‑date – Skills pull the latest definitions from the repo; no manual version pinning required.
• Enterprise‑ready – Works with Scenario’s authentication and can be tied to team accounts for asset governance.

Limitations

• Dependency on Scenario MCP – All generation happens in the cloud; offline use is not possible.
• Learning curve – Users must understand the Agent Skills syntax and how to configure the skills CLI.
• Expert‑tool families – The optional DCC integrations require local installations of heavy software (Blender, Unreal, etc.) and are not covered by the core MCP flow.

---

Who Should Use This?

• Game developers who want AI‑generated art, textures, and 3‑D props directly from code.
• Marketing teams needing rapid on‑brand visual production (product shots, ad variants, thumbnails).
• Content creators looking to automate video, music‑video, and audio generation.
• LLM‑tooling engineers building custom agents that can request visual/audio assets without writing bespoke API code.

---

Getting Help

• Documentation – API docs (https://docs.scenario.com), MCP docs (https://mcp.scenario.com/docs), and skill‑specific READMEs.
• Community – Follow @scenario_gg on X, and browse the skill index at https://skills.sh/scenario-labs/skills.
• Support – Sign in to the Scenario app and use the help center for account or API‑key issues.

---

TL;DR

Scenario Agent Skills is a library of 71 plug‑in definitions that let AI coding assistants talk to Scenario’s cloud asset‑generation platform. Install the desired subset with npx skills add, connect to your Scenario MCP endpoint, and then prompt your LLM to create images, videos, audio, 3‑D models, or full brand kits—all without writing any API code yourself.