Agent Skills that teach AI coding agents (Claude Code, Cursor, Codex, Copilot, and 70+ others) how to create production-ready content with Scenario through the Scenario MCP server: images, video, audio, textures, skyboxes, 3D assets, and custom-trained models, for games, entertainment, and any creative vertical.
Skills follow the Agent Skills format.
# Every skill, no picker
npx skills add scenario-labs/skills --skill "*"
# Pick from the installer menu (it opens with nothing preselected)
npx skills add scenario-labs/skills
# A single skill
npx skills add scenario-labs/skills --skill scenarioSome skills direct sibling skills per stage, so installing one of them alone leaves its references unfilled. Install a composed pipeline together:
# Video ads (director skill plus the siblings it delegates to)
npx skills add scenario-labs/skills --skill scenario --skill scenario-video-ads --skill scenario-image --skill scenario-video --skill scenario-seedance --skill scenario-audio --skill scenario-video-assembly --skill scenario-text-overlay --skill scenario-consistency --skill scenario-asset-analysis
# Seedance music video
npx skills add scenario-labs/skills --skill scenario --skill scenario-seedance-music-video --skill scenario-seedance --skill scenario-consistency
# Seedance storyboarded choreography
npx skills add scenario-labs/skills --skill scenario --skill scenario-seedance-storyboard --skill scenario-seedance --skill scenario-consistency --skill scenario-video-editing --skill scenario-video-assemblySkills need the Scenario MCP server connected:
claude mcp add --transport http scenario https://mcp.scenario.com/mcpOr add https://mcp.scenario.com/mcp to any MCP client and sign in with a Scenario account (OAuth), or use an API key. Full setup lives in the scenario skill.
Installs always fetch the latest copy of each skill from main; there is no version pinning. Releases are how changes become visible: every feature, fix, and documentation change merged to main lands in the next release, staged by release-please as an open release PR. Merging that PR tags a version, updates CHANGELOG.md in this repository, publishes a GitHub release, and posts the notes to the Scenario changelog.
- To learn that a skill changed, watch this repository's releases (Watch, then Custom, then Releases on GitHub) or check any of the changelog pages above. An entry scoped to a single skill carries that skill's name; changes spanning several skills appear under broader scopes such as
skills. - To pull the changes, run
npx skills updatewhere you installed, or re-run your originalnpx skills addcommand.
Connect the Scenario MCP server and learn the core generation loop every other skill builds on.
| Skill | Use it for |
|---|---|
| scenario | Connecting to the Scenario MCP and the core generation loop: discover, schema, run, wait, display, download |
Turn a blank brief into references, options someone can choose between, and a moodboard the next batch runs from.
| Skill | Use it for |
|---|---|
| scenario-inspiration | Finding a direction before generating: serendipity, reference hunting, A/B/C/D concept options, moodboard collections |
Generate and edit images: model choice, sizing, references, masked edits, post-processing tools, and letter-perfect text overlay cards.
| Skill | Use it for |
|---|---|
| scenario-image | Text-to-image and image editing: model choice, sizing fields, prompt limits, reference images, masked inpainting |
| scenario-product-shots | Product photography from a real photo: packshots, lifestyle scenes, relighting, label-fidelity gates |
| scenario-brand-kit | Visual identities: SVG logos and wordmarks, palette and type specs, gated variants and applications, the filed kit |
| scenario-image-editing | Tool-model image edits: 3D LUT grades, effects, expand and reframe, resize, slicing, layers, background removal |
| scenario-text-overlay | Letter-perfect text overlays: templated transparent PNG cards (taglines, CTAs, legal supers, rich cards) to composite |
| scenario-storyboards | Comic pages, storybooks, and pre-viz storyboards: script first, one run per panel, a locked cast, lettering in post |
Sprites, icons, tilesets, textures, skyboxes, and 3D assets ready for game engines.
| Skill | Use it for |
|---|---|
| scenario-game-assets | Sprites, icons, props, tilesets, pixel art, concept art, transparent backgrounds, style-consistent batches |
| scenario-sprite-animation | Walk cycles, idle loops, VFX sprites, and animation sheets from one prompt: GIF vs sprite sheet output, frame slicing, pixel cleanup |
| scenario-textures | Seamless and tileable textures, PBR materials, tiling-safe upscaling, engine-ready sizing |
| scenario-skyboxes | 360 equirectangular panoramas and skyboxes, seam-safe upscaling, engine export |
| scenario-3d | Text or image to 3D meshes, multi-view reconstruction, retexture and remesh, inline 3D preview, GLB/FBX download |
| scenario-patina-retexture | PATINA PBR retexture of a finished mesh: material families, one map set per family, Blender apply with geometry untouched, matched before/after comparison film |
| scenario-orbit-views | New camera angles of one picture through a 3D intermediary: clay layouts in a grounded panorama, repaint per camera, matched transparent set |
Video generation and editing, music, sound effects, voice, and speech.
| Skill | Use it for |
|---|---|
| scenario-video | Text-to-video and image-to-video, motion prompting, lipsync, video editing, upscaling, cut/split/concat utilities |
| scenario-video-editing | Tool-model footage edits: LUT grades and effects, trim, split, resize, reverse, frames, masks, layers |
| scenario-seedance-music-video | Turning a song into a music video: beat-aligned shots, lyric transcription, shot sound under the master, assembly |
| scenario-seedance-storyboard | Movement that holds across cuts: timecoded shot scripts, character sheets and boards, pose-chained shots that play as one performance |
| scenario-video-ads | Producing a video ad from a product shot: brief, storyboard, cinematic grammar, fidelity gates, budget, delivery |
| scenario-ugc | UGC-style creator video: talking-head and avatar ads, demos, faceless voiceover, spoken-register scripts, claim safety |
| scenario-fan-cam | Personalized fan-cam clips: identity edit into a broadcast still, reaction beats to video, graphics composited in post |
| scenario-audio | Music, sound effects, voice and speech generation, video scoring, audio utilities |
| scenario-video-assembly | Assembling clips into a finished video: timeline composition, concat with transitions, overlays, music, captions |
| scenario-caption-studio | Captioning a finished video per destination: styled burn-in or SRT, transcription hints, translation, per-platform placement |
Hold one character, product, or style across a whole set, and train custom models on your own art.
| Skill | Use it for |
|---|---|
| scenario-consistency | Holding one character, product, or style across a set: baseline-plus-delta prompts, reference images, control maps |
| scenario-identity-library | Creating named characters and props as reusable identities: interview to brief, gated collections, Grid Maker sheets, reuse |
| scenario-model-training | Training custom models for style, character, or product consistency, and generating with them |
Read finished assets back: caption them, extract a style or a control map, check a batch against a brief or the Quality Gate, compare candidate models on one brief, and file the keepers.
| Skill | Use it for |
|---|---|
| scenario-asset-analysis | Reading assets back: captions, style descriptions, batch review against a brief, control maps, collections, tags |
| scenario-quality-gate | Pass/warn/fail image verdicts from the Quality Gate: free stored reads, dry-run pricing, feeding suggestions back into the next run |
| scenario-refine-loop | Iterating until output matches the brief: rubric first, batched critique verdicts, cheapest targeted fix, round caps |
| scenario-model-comparison | Bake-offs between candidate models on one brief: normalized inputs, a dry-run cost matrix, billed cost and latency, contact sheets, pre-registered criteria |
Ship one approved master, image or video, to every placement: ratios, safe zones, and platform specs from social feeds to shops, storefronts, and print.
| Skill | Use it for |
|---|---|
| scenario-formats | Deriving every placement from one master, image or video: crop vs resize vs expand vs reframe, platform specs and safe areas |
Discover, run, build, and publish Scenario workflows (apps), and migrate graphs from Weavy, ComfyUI, or other node editors.
| Skill | Use it for |
|---|---|
| scenario-workflows | Running saved workflows (apps): building the run inputs, dry-run pricing, unsticking approval gates |
| scenario-workflow-authoring | Creating and editing workflow graphs: editor_info grammar, node wiring, publishing apps, and migrating graphs from Weavy, ComfyUI, or other node editors |
Get past a blocked generation, and turn a stuck session into a reproducible public issue.
| Skill | Use it for |
|---|---|
| scenario-moderation | Recovering a blocked generation: provider-side filters, model switching, proportional wording, look-alike errors |
| scenario-report | Reporting a bug or change request as a reproducible, redacted issue on this repository's public tracker |
Govern a team from the agent: which models it can run, who is on it and in which projects, per-member spend caps, API key roles, and who spent what.
| Skill | Use it for |
|---|---|
| scenario-team-admin | Model access control, team and project membership, per-member Creative Unit caps, API key roles, and spend attribution |
| scenario-admin-analytics | Cached MCP usage analytics, model and user rankings, CSV exports, adoption PDFs, and a local dashboard |
The top image families in depth: generation and editing modes, reference rules, and the caps that differ per member.
| Skill | Use it for |
|---|---|
| scenario-gpt-image | GPT Image 2.5 Flare and Sunburst, GPT Image 2 and 1.5: member routing, the quality tiers up to max and their price ladder, alpha masks, transparent cutouts, in-image text, pixel sizing |
| scenario-mai-image | MAI Image: in-image typography, quoted copy prompts, instruction editing (preserve first, one change), per-editor source input shapes, edit cost split |
| scenario-seedream | Seedream images: member selection (Pro, Lite, 4.5, Layerize), per-member sizing contracts, in-image text, sequence sets, layer splits, cost |
| scenario-gemini-image | Gemini image (Nano Banana): member choice (Flash, Pro, Lite), instruction editing, reference roles, video stills, Search grounding, cost |
| scenario-reve | Reve image: v2.1 vs Remix selection, frame-tag reference wiring, maskless instruction edits, per-member caps and cost |
| scenario-ideogram | Ideogram images: member selection (V4 typography, native transparency, layerize text, character, remove background), expansion off for exact copy, transparency routes, per-member parameter drift |
| scenario-grok-imagine-image | Grok Imagine images: text-to-image and instruction editing, exact in-image typography, quoted copy rules, aspect ratio traps, quality and cost tiers |
| scenario-luma-image | Luma Uni-1 images: create vs edit mode, role-labeled references, web search grounding, rendered text, cost |
The top video families in depth: conditioning modes, native audio, editing and extension, and per-member caps and costs.
| Skill | Use it for |
|---|---|
| scenario-minimax-video | MiniMax Hailuo video: H3 keyframe vs reference modes, bracketed camera commands, native audio prompting, 2.3 duration coupling, cost |
| scenario-gemini-omni | Gemini Omni video: member selection (first frame, references, edit, extend), prompted native audio, identity from references, motion-locked restyles |
| scenario-grok-imagine-video | Grok Imagine video: one member per mode (first frame, @image references, edit, extend), prompted audio and dialogue, resolution and preprocessing caps |
| scenario-veo | Veo video: mode selection (first frame, transitions, references, extend), asset or style references, prompt-directed native audio, tier costs |
| scenario-seedance | Seedance video: mode selection (first frame, references, edit, extend), conditioning traps, native sound, cost |
| scenario-kling | Kling video: picking a member across V3, O1, and 2.6 (multi-shot, elements, editing, motion control, lipsync, avatar), audio exclusivity, tier cost |
| scenario-vidu | Vidu video: tier and mode selection (text, image, start/end frames, references), per-tier caps, two prompt shapes, music toggle traps |
| scenario-wan | Wan video: member-per-job selection (T2V, I2V, edit, Animate, reframe, outpainting), parameter drift across generations, multi-shot prompts, audio sources |
| scenario-runway | Runway video: Gen4.5 generation (text, first frame), Aleph 2 footage edits (swap, remove, restyle), keyframe pinning, aspect ratio traps, cost |
| scenario-luma-video | Luma Ray video: member routing (generate, edit, reframe, modify), option vetoes (10s, loop, HDR, anchors), twin editor ladders, cost |
The top audio families in depth: music, covers and stems, sound effects, speech, dubbing, and video scoring.
| Skill | Use it for |
|---|---|
| scenario-minimax-music | MiniMax music: vocal songs from a tagged lyric sheet, auto lyrics, instrumental mode, melody-keeping covers, character-count cost |
| scenario-elevenlabs | ElevenLabs audio: picking the member (TTS, music, SFX, dubbing, re-voicing, isolation), voice and tag rules, song section grammar, cost |
| scenario-ace-step | ACE-Step music: one member per mode, Turbo vs Quality lanes, lyric sheet tags, cover strength, repaint windows, stem edits |
| scenario-sonilo | Sonilo audio: SFX and music from text or video, standalone track vs muxed clip, segment prompts, keep-speech scoring, cost |
The top 3D families in depth: mesh generation, retexture and rigging toolchains, portraits, and explorable worlds and splats.
| Skill | Use it for |
|---|---|
| scenario-meshy | Meshy 3D: generate from image, multi-view, or text, then retexture, remesh, UV unwrap, rig, or animate; texture precedence, Ultra mode, polycount traps |
| scenario-rodin | Rodin 3D: member selection (image-to-3D, text-to-3D, Fast lanes, Bang! mesh splitting), full versus Fast parameter dialects, topology and tier traps, cost |
| scenario-sparc3d | Sparc3D image-to-3D: ordered multi-view input, mesh vs textured output, per-member resolution spellings, face budgets, long-job waits |
| scenario-3d-worlds | 3D worlds and splats: member choice by input (text, image, pano, multi-view, video, object), Marble draft-then-upgrade seed reuse, HY World skybox-to-splat pipeline, long-job patience |
Once the skills are installed and the MCP server is connected, ask your agent things like:
- "Generate four style-matched potion icons with transparent backgrounds for my RPG inventory"
- "Make a seamless brick texture, then upscale it to 2048 without breaking the tiling"
- "Turn this concept sketch into a 3D prop and let me preview it before I export the GLB"
A skill is a SKILL.md file with procedural knowledge an agent loads on demand, defined by the open Agent Skills specification. The skills CLI installs these files into 70+ agents, and the ecosystem directory lives at skills.sh.
See CONTRIBUTING.md for the contribution workflow, and AGENTS.md for the authoring contract, the public-content policy, validation, and the application-testing bar. One-time setup after cloning: pnpm install (installs commitlint, cspell, prettier, and the husky git hooks). pnpm run validate runs the same content checks CI runs; commit messages and the PR title are linted separately with commitlint. PRs welcome; PR titles follow Conventional Commits since they become the squash commit header on main.
Every script shipped with a skill has a test suite in tests/<name>/; pnpm test runs them all (CI does too). Suites need Python 3.11+ with any tests/<name>/requirements.txt dependencies installed, plus ffmpeg on PATH where a suite uses it. Python suites use stdlib unittest, TypeScript suites use vitest.