SkillFoxx

Flovart skill

A skill that teaches a coding agent to work on the Flovart canvas: build generation nodes, run them with consent.

  • 145 ★ on the repository
  • Last commit
  • MIT
Skill

Extended access

Why this level

  • Generation goes through paid provider APIs and is billed on their side.
  • The agent changes the project on the canvas and runs nodes through the local CLI.
All reasons and checks

avabbbb/flovartAuthor:avabbbb

Install

Manual install

git clone https://github.com/avabbbb/Flovart.git
cd Flovart
npm install
npm run flovart:cli -- start --source --web --open
npm run flovart:cli -- status --json

Run from source while the versioned CLI package stays under release control. The skill is in `.agents/skills/flovart/`, with a Claude copy in `.claude/skills/flovart/`. Node.js 22.19.0 or newer is required.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Generation goes through paid provider APIs and is billed on their side.
  • The agent changes the project on the canvas and runs nodes through the local CLI.

This is third-party code. Review the repository files before installing.

What it does

Flovart is a local visual canvas with a workflow for generating images and video. The `flovart` skill explains to a coding agent how to work on that canvas while the author sees every change and can edit by hand. The agent calls the local `flovart` CLI and uses five operations: `status`, `workflow.inspect`, `workflow.selection.get`, `workflow.apply` and `workflow.node.run`. `ensure` and `doctor` handle preparation and diagnostics. Before a paid generation the skill requires naming the number of runs, the model and the purpose of each, then waiting for consent. The agent neither asks for nor reads provider keys: `provider.status` returns only masked values. Flovart bundles no model services; you connect a provider in the app settings with your own key.

Who it is for. Creators and video editors who make images and clips on the Flovart canvas and want an agent to assemble nodes and run generations.

Good fit when

  • A storyboard is built from several generation nodes, and the agent lays them out on the canvas while the author watches.
  • Dozens of drafts run as a batch through the table, and before launch the agent states the number of paid calls.
  • A result must be compared with the original: the agent creates candidates, and a person chooses.
  • The agent attaches to an already open canvas and checks the project title and node count before editing.

Not a fit when

  • A stable product is required: Codex CLI with Browser Workflow is Beta, while Claude Code and OpenCode are Experimental.
  • Node.js 22.19.0 or newer is missing: the CLI will not start.
  • Generation must run without paying a provider: Flovart bundles no models, and paid runs go through your key.

Example request

Open Flovart and work on this workflow: add three generation nodes from my brief, state the number of runs and the model, and wait for confirmation.

Limitations

The product is in closed beta 0.2.0: Windows and macOS builds are unsigned, so SmartScreen and Gatekeeper ask for confirmation. The versioned CLI package is still passing release control, and the skill is run from source through `npm run flovart:cli --` rather than `npx flovart-cli`. Integrations with Claude Code, OpenCode and DaVinci Resolve are marked experimental; certification covers one paid provider, RunningHub. Paid generations are billed on the provider side even if you stop a run. Some error messages are still Chinese-only.

How to disable. Delete the skill directory `.agents/skills/flovart/` or `.claude/skills/flovart/` and stop the local Flovart service. Projects and generated results stay on disk and are deleted separately.

FAQ

How do I connect Flovart to an agent?

Clone the repository, run `npm install` and `npm run flovart:cli -- start --source --web --open`. On a source checkout the agent finds the skill in `.agents/skills/flovart/` by itself, and Claude has a copy in `.claude/skills/flovart/`.

Which operations does the agent have?

Five: `status`, `workflow.inspect`, `workflow.selection.get`, `workflow.apply` and `workflow.node.run`. `ensure` and `doctor` handle connection and diagnostics.

Does the agent see my provider keys?

No. `provider.status` returns the list of connected services with masked keys, and the skill forbids asking for a key in the conversation.

Who pays for generation?

You do, to the chosen provider: Flovart bundles no model services. The skill requires stating the number of runs and the model and waiting for consent.

Which Node.js version is needed?

Node.js 22.19.0 or newer. If the `flovart` command is not installed, the skill runs the CLI from the repository through `npm run flovart:cli --`.

Discussion

No questions or reviews yet.

Skills

The official Lovart skill: an agent draws images, makes video, music and 3D models and keeps projects with chat history

SkillExtended access149

Flue gives a coding agent shell access to the scripting APIs of Photoshop, Blender, Unity, Office and other desktop software.

SkillBroad access108

Two skills for generative video and images: video writes Seedance, Kling and Veo prompts by dramaturgy rules, image picks the model and its syntax

SkillMinimal access658

A skill that writes prompts for video, image and music generation on 14 platforms and can submit them in the browser

SkillBroad access274

A skill and CLI that let Claude Code or Codex make videos as web pages, with the Studio editor to change the result

SkillExtended access5.9K

Kling CLI

kling-cli

Official

A skill for the Kling command-line tool: the agent makes video and images from text and a reference, drives motion and keeps reusable characters

SkillExtended access1

README in short

The documentation presents Flovart as a local workflow tool for people and agents with Canvas, Table and Agent surfaces. Next come the beta status, a compatibility matrix with Codex CLI, Claude Code, OpenCode and creative apps, and a roadmap around DaVinci Resolve Studio.

README badge

Original SKILL.md

---
name: flovart
description: Work in the user's visible Flovart Workflow canvas. Use when the user asks to open Flovart, look at or change their Workflow, build shots or generation nodes, or run image and video generation.
---

# Flovart

Flovart is the visible Workflow canvas the user is looking at. You change it
through the local `flovart` CLI; the user sees every change live. Flovart
prepares its own local session: never ask the user for connection details and
never rebuild a connection flow in the conversation.

The CLI needs Node.js 22.19.0 or newer. In this source checkout, when the
packaged `flovart` command is not installed, use the repository entrypoint
instead, for example `npm run flovart:cli -- ensure --json`. Do not use `npx flovart-cli` as an automatic fallback: an unpublished checkout can make
npx fetch an unrelated package.

## Prepare for this job

Do these in order at the start of every task. Each step has a stop condition;
never skip ahead. Details are in `references/environment.md`.

1. **Understand the request.** Decide what the user wants to see on the canvas
   when you are done, and whether it needs paid generation. Ask only when a
   different reading would change the result.
2. **Make Flovart ready.** Run `flovart ensure --json`. Continue only when it
   returns `ok: true` and `state: "ready"`; otherwise tell the user the public
   state it returned and stop. `ensure` also opens the user's browser on the
   canvas when needed.
3. **Get on the user's canvas.** Run `flovart status --json`, then
   `flovart workflow.inspect --agent-identity <host> --json`. Compare the
   project title, node titles and node count with what the user describes or
   shows you. If they disagree, you are reading another canvas: stop and fix
   that first (see "Operate the user's visible canvas").
4. **Check AI services only when generation is needed.** Run
   `flovart provider.status --json`. It lists configured services with masked
   keys only. If the route you need is missing, tell the user where to add it
   in Flovart Settings (`references/providers.md`). Never ask for a key in the
   conversation, and never read, write or move one yourself.
5. **Quote and agree.** Before any paid run, tell the user how many runs, which
   model, and what each run is for, and wait for their go-ahead.
6. **Work and narrate.** Make the change, inspect again, and tell the user what
   is now on their canvas in their terms.

## The stable surface

Use `ensure` to prepare, then these five operations for all normal work:
`status`, `workflow.inspect`, `workflow.selection.get`, `workflow.apply`, and
`workflow.node.run`.

Report a problem