Pruna Skills
Pruna Skills teach your coding agent how to generate images, video, and audio with the Pruna AI API. They also cover multi-step workflows — explainers, music videos, avatars, illustrated reels — that chain those calls together. Each skill follows the Agent Skills format, so it works in Cursor, Claude Code, Copilot, Codex, and many other agents.
Quickstart
Set your API key, install the /pruna suite skill, and start a new chat.
export PRUNA_API_KEY="your_key"
npx skills add PrunaAI/pruna-skills@pruna -y
See docs/api-setup.md for key setup.
That installs the full suite — every guide, tool, and workflow in one shot. After install, open a new chat and describe what you want in plain language — your agent picks the right skills from the suite.
Try it: ask your agent
Create an image, then try on clothes, then create a video
Create a portrait of a teenage girl drummer in a garage. Put a vintage red band jacket on her from a garment reference, and keep the pose and background. Then make her sing along to a song slice as a lip-sync performance clip.
Step 1 · p-image | Step 2 · p-image-try-on | Step 3 · p-video-avatar |
|---|---|---|
| Image | Try-on | Video |
 |  |  |
Create an image, then edit it, then create a video
Create a product photo of a white running sneaker on a shelf. Edit it to the same shoe in bright orange. Animate a color shift, then a hand picking it up.
Step 1 · p-image | Step 2 · p-image-edit | Step 3 · p-video |
|---|---|---|
| Image | Edit | Video |
 |  |  |
Create an image, then add narration, then assemble a video
Create an illustrated story image of a whale in a library, narrate a short line, and assemble a Ken Burns reel.
Step 1 · p-image | Step 2 · gemini-3.1-flash-tts | Step 3 · assembly |
|---|---|---|
| Image | Narration | Video |
 | MP3 link above |  |
More examples: docs/EXAMPLES.md. Ready-made multi-step recipes: docs/WORKFLOW-RECIPES.md.
How it works
Skills come in four types:
| Type | What it does | Example |
|---|---|---|
| Suite | Everything in one install | pruna |
| Guide | Prompting craft and API safety — no generation call on its own | image-prompting |
| Tool | One paid endpoint — a single Pruna or Replicate API call | p-image |
| Workflow | A multi-step playbook your agent runs with curl and ffmpeg | music-video |
Guides teach how to prompt and call the API safely. Tools are the endpoints — one model per skill (p-image, p-video, music-2.5, …). Workflows chain tools into finished deliverables.
When you install pruna, you get the whole catalog. Installing a single tool instead pulls in only the guides listed under its Prerequisites. Workflows list the tools they need — you only install those separately if you skipped the suite.
Install one skill at a time
Prefer to start small? Pick a single skill:
npx skills add PrunaAI/pruna-skills@p-image -y # one tool (+ its guides)
npx skills add PrunaAI/pruna-skills@music-video -y # one workflow
npx skills add PrunaAI/pruna-skills -l # list all
What's in the suite
The pruna suite includes 26 skills across guides, tools, and workflows. You don't need to memorize them — describe your goal in chat and your agent picks what fits.
- Browse by goal: docs/SKILL-CATALOG.md
- Per-skill install commands: skills/suite/pruna/SKILL.md
- Multi-step recipes: docs/WORKFLOW-RECIPES.md
API setup
Set PRUNA_API_KEY for images and video. Audio tools also need REPLICATE_API_TOKEN. Step-by-step instructions: docs/api-setup.md.
Other install channels
ClawHub and related channels are supported too. See PUBLISHING.md.
Contributing
Installable skills live under skills/ (guides, tools, workflows, and the pruna suite). Human docs live in docs/. Maintainers: make bundle && make validate — see CONTRIBUTING.md.
License
See LICENSE.