Runware Skills

Ask for the result,
not the model.

Packaged recipes that teach your agent to get the best results from Runware. And keep getting better as the models improve.

Get an API KeyBrowse the catalog
OPEN SOURCE · LOAD ON DEMAND · RUNS IN YOUR AGENT
Routing flow
  1. You ask

Why skills

The skill stays.
The model improves.

Ask for the outcome once. The skill picks the model that fits, and upgrades to a better one as the catalog moves, with no change on your side.

Skill

character-consistency

You ask

Put this character in a new scene, keeping the same face.

Reference photo of the character used in the skill demo
Now

Nano Banana 2 Lite

Character in a new scene generated with Nano Banana 2 Lite
Future model

Future model

Same character in a new scene generated with a future model

Same skill. Always the model that fits.

The catalog

A skill for the work you actually do.

Three foundation skills carry the shared machinery. Every outcome skill builds on them and adds the craft for one specific job.

Agent skills for image, video, audio and 3D

30+ skills across every modality

Foundation

Shared machinery every outcome skill builds on.

  • runware-run

    Send schema-valid fields and return the result.

  • runware-models

    Pick the right model and keep that choice current.

  • runware-prompting

    Per-model prompt craft.

  • product-photographyPackshot into studio, lifestyle, hero shots.
  • character-consistencyKeep a character or product identical across scenes.
  • text-in-imagePosters, packaging, UI with exact copy.
  • edit-imageAdd, remove, replace, recolor, relight, inpaint, outpaint.
  • restore-and-upscaleDeblur, denoise, upscale, restore old photos.
Browse the full catalog
How it works

You ask. The right skill answers.

A skill is a small folder with a SKILL.md. Its description is the trigger: when your request matches, the agent reads the skill and follows it.

Loads on demand

Only a skill's name and description sit in context until your request matches. Install the whole catalog, pay almost nothing until one runs.

Never hardcodes a model

A skill names a default, confirms it's live, and reads its current schema before running, so it stays correct as the catalog changes.

Skills combine

When one request needs more than one outcome, the matching skills run together. Ask for a video with a voiceover, and you get both from a single sentence.

How skills run

Skills are the know-how.
Your agent does the work.

Use skills when you want a result without scripting the steps. They run two ways: on their own inside an agent, or in a conversation with an AI chat.

In an agent

Runs as part of the job

An agent or workflow picks up the skill and runs it automatically, no prompting step by step.

In an AI chat

Runs in the conversation

Ask in plain language in a chat like Claude, and it runs the skill and hands back the result.

Install

Add the ones you want.

Add the skills your agent needs. Each one loads on demand.

/plugin marketplace add runware/runware-skills
/plugin install skills@runware

or clone the repo and copy folders into .claude/skills/

Models across every modality

Every new model.
The day it ships.

New models land on Runware the day they ship, and your skills can use them right away.

See full model library
FAQ

Questions, answered.

Packaged recipes that let an agent get good results from Runware without knowing the catalog.

Runware Skills

Give your agent the recipes.

Browse the catalog, add what you need, ask in plain language.

Get an API keyRead the docs