lovart-skill
The official Lovart skill: an agent draws images, makes video, music and 3D models and keeps projects with chat history
- 149 ★ on the repository
- Last commit
Extended access
Why this level
- The AK/SK keys open access to a Lovart account, and every generation spends paid credits.
- Prompts and uploaded references go to an external service, so confidential mockups should not be sent there.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/lovart-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The AK/SK keys open access to a Lovart account, and every generation spends paid credits.
- Prompts and uploaded references go to an external service, so confidential mockups should not be sent there.
This is third-party code. Review the repository files before installing.
What it does
lovart-skill connects an agent to the Lovart open API. One phrase such as 'draw a cyberpunk cat in a neon city' becomes a `chat` call in the `agent_skill.py` script: the skill waits for generation to finish, downloads the result and returns the files. It handles images (posters, logos, illustrations, banners), video, audio (background music, songs, sound effects), 3D models from text or an image, and edits: upscaling, reframing and style transfer. A request can be steered to a particular model: name it in the prompt ('video with Kling'), softly through `--prefer-models` or strictly through `--include-tools`; a catalog of dozens of image and video models is available. The `--mode thinking` option adds planning for complex jobs such as designing a brand identity. Projects and chats are stored on the Lovart server and locally in `~/.lovart/state.json`: the `projects`, `project-switch` and `threads` commands return to an earlier thread, where a result is refined with a phrase like 'make it blue'. Expensive operations such as video stop at a confirmation: the agent names the approximate cost in credits and waits for an explicit yes.
Who it is for. Designers, marketers and content authors who hand an agent the preparation of graphics, clips and music in an already paid Lovart service.
Good fit when
- The agent makes a poster, banner or logo from a description and saves the files in the project folder.
- A short clip is needed for a post, and a person confirms the expensive video generation.
- A presentation needs background music or a sound effect without searching libraries.
- A finished picture must be upscaled, reformatted for a social network or restyled after another work.
- A brand is worked on in several rounds, and each round returns to the earlier project and thread.
Not a fit when
- You have no Lovart account and AK/SK keys: the skill does not work without them.
- Local generation is needed with no prompts or files sent to the cloud.
- Spending credits on every iteration is not an option: expensive operations and repeats of a rejected reference use up the balance.
Example request
Draw a retro-style poster for a coffee shop, then make a vertical version of it for stories.Limitations
The keys `LOVART_ACCESS_KEY` and `LOVART_SECRET_KEY` from the avatar menu on the Lovart platform, Python 3 and a credit balance are required. Only the OpenClaw version is officially published through ClawHub, and for Hermes Agent the skill is copied by hand. A generation can finish with status done although some tool calls were rejected, so the `failures` array in the answer must be read. A retry with a rejected reference fails again and still spends credits. The `--exclude-tools` flag is accepted but has no effect on tool choice yet.
How to disable. Delete the skill folder from the agent skills directory and clear `LOVART_ACCESS_KEY` and `LOVART_SECRET_KEY`. Local state sits in `~/.lovart/state.json`, and the keys can be revoked in the AK/SK menu of the Lovart platform.
FAQ
How do I get keys for lovart-skill?
Open the avatar menu on the Lovart platform, the AK/SK Management section, and create a key pair. Export them as `LOVART_ACCESS_KEY` and `LOVART_SECRET_KEY`.
How do I install the skill in OpenClaw and Hermes Agent?
For OpenClaw run `npx skills add lovartai/lovart-skill`: the release comes from ClawHub. For Hermes Agent clone the repository and copy `skills/lovart-skill` to `~/.hermes/skills/design/lovart-api`.
Will credits be spent without my consent?
For expensive operations such as video, the skill stops at the pending_confirmation status, names the approximate cost and waits for an answer. The agent must not confirm the charge itself.
How do I choose a model for generation?
Name the model in the request, pass `--prefer-models` as a soft wish or `--include-tools` as a strong hint. The latter does not guarantee the choice: after a rejection the agent may pick another tool.
What if the result contains failures?
Read the array and the warning: some calls were rejected, and the agent may have dropped a reference or switched models. A retry with the same input fails again and spends credits, so change the input first.
Discussion
No questions or reviews yet.
Related
SkillsA skill that teaches a coding agent to work on the Flovart canvas: build generation nodes, run them with consent.
A skill for the LibTV platform: from a text brief an agent makes images and video, edits finished clips and downloads the result.
Five image-generation skills: Qwen-Image prompts, logos with shared geometry, diary comics, transparent PNGs and process dioramas
Two skills for generative video and images: video writes Seedance, Kling and Veo prompts by dramaturgy rules, image picks the model and its syntax
The agent scripts, voices and draws a narrated documentary in code, with one continuous camera over photos, maps and cards.
Thirty-nine skills: video editing, landing pages from a clip, decks, AWS, Vercel, Cloudflare, Meta Ads and OWASP
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| lovart-skill (this card) | no | 149 | |
| Higgsfield CLI | yes | 711 | the official Higgsfield console with image, video, 3D and audio models; this is the Lovart skill with projects, chat threads and cost confirmation |
| Kling CLI | yes | 1 | a skill for the Kling console for video and images; here the request goes to the Lovart agent, which picks the tool itself, and audio with 3D are added |
README in short
The documentation gives two install paths, command lists for generation, projects, settings and files, call examples and a table of models. Separate sections explain model choice, reference images from the library, reasoning modes and handling rejected calls. MIT license.