beautiful-feishu-whiteboard: 35 Curated Palette Styles for Feishu Agent Whiteboards

A Palette and Constraint Library for AI-Generated Feishu Whiteboards

beautiful-feishu-whiteboard is an agent skill for Claude Code and other agents that load SKILL.md. It solves a specific problem: AI agents composing Feishu or Lark whiteboards need both tasteful color choices and accurate knowledge of what the Feishu SVG whiteboard renderer supports. Without a curated palette, agents make arbitrary color decisions. Without RULES.md, they attempt shapes, opacity values, and text color configurations that the Feishu renderer rejects or mishandles.

The repository is not an auto-layout chart tool. The agent decides how to arrange shapes and content; beautiful-feishu-whiteboard provides the palette, mood, and hard rendering rules. The output is a real, editable Feishu whiteboard embedded in a Feishu doc, not a static image or screenshot. Team members opening the doc can move shapes, edit text, and apply further changes directly inside Feishu.

Thirty-Five Styles Ordered from Restrained to Bold

The 35 styles are organized into three groups that the README calls Restrained, Balanced, and Bold. The Restrained group contains nine styles including Avocado Press, Grove, Jade Lens, Long Table, Macchiato, Monochrome, Papier Bleu, Reading Room, and Salmon Stamp. These use subdued color relationships and minimal contrast. The Balanced group has fifteen styles including Apricot Arc, Berry Pop, Bold Poster, Checker Bloom, Cobalt Bloom, Coral, Cut Bloom, Editorial Forest, Lime Slab, Linen Cut, Pin and Paper, Raw Grid, Riptide Cobalt, Soft Editorial, and Violet Marker. The Bold group has eleven styles including BlockFrame, Burst Panel, Confetti Wedge, Court Press, Crayon Stack, Grove Block, Mint Brut, Neo-Grid Bold, Riso Brut, Specimen Bold, and Stencil and Tablet.

The gallery in the README renders all 35 styles with identical content, specifically the three stages of LLM training, so a user comparing styles sees the same shapes and text in each palette. This makes direct comparison of color decisions practical. Each style is a template in the templates/ directory, accessible by clicking the style name in the README.

Installing the Skill and Setting Up lark-cli

There are three installation methods. The first uses the npx skills installer at project scope:

npx skills add zarazhangrui/beautiful-feishu-whiteboard

For global installation, available in every project:

npx skills add zarazhangrui/beautiful-feishu-whiteboard -g

The third method is manual, cloning directly into the agent's skills folder:

git clone https://github.com/zarazhangrui/beautiful-feishu-whiteboard \
  ~/.claude/skills/beautiful-feishu-whiteboard

Before using the skill to publish a whiteboard, lark-cli must be installed and authenticated. The README lists the setup steps:

npm install -g @larksuite/cli
lark-cli config init     # first run: scan the QR code
lark-cli auth login      # authorize your Feishu/Lark account

The lark-cli config init step opens a QR code for the first-time scan. Node.js 20 or newer and an active Feishu or Lark account are required before this sequence will succeed.

RULES.md: What the Feishu SVG Renderer Can and Cannot Do

The core technical contribution of the repository is RULES.md, not the color choices alone. The README describes it as encoding hard-won, on-board-verified knowledge of the Feishu SVG whiteboard renderer. The specific constraints listed in the README's description include: native shapes only, no opacity, and a text color export quirk. These are not documented by Feishu's own developer documentation.

For an agent generating Feishu whiteboards without RULES.md, these constraints are invisible until the whiteboard fails to render correctly or exports with wrong text colors. The RULES.md file captures the boundary between what the renderer will accept and what it will silently mishandle, derived from direct testing on the Feishu platform. This makes the repository useful even for teams that do not need any of the 35 color palettes: RULES.md alone is a reference for building custom whiteboard generation tools against the Feishu SVG renderer.

The agent entry point is SKILL.md at the repository root. Agents that load this file receive both the palette instructions and the rendering rules together. The templates/ directory holds one subdirectory per style, each containing the palette definition for that style. CATALOG.md in the repository root provides an index of all 35 styles. The assets/ directory holds supporting image assets for the gallery, and the scripts/ directory contains the Shell-based utilities included in the repository.

Where beautiful-feishu-whiteboard Does Not Help

The skill does not compose layouts. An agent instructed only to use this skill will still need to decide where to place shapes, how to label them, and how to structure the whiteboard's content. The palette and rules give visual consistency; the spatial and logical arrangement of information remains the agent's responsibility.

The skill works only within Feishu and Lark. It has no equivalent for Miro, Figma, Confluence whiteboards, or Excalidraw. Teams whose collaboration platform is not Feishu or Lark get no value from this repository.

The npx skills installation path requires Node.js 20 or newer. Environments running older Node.js versions need the manual git clone approach. The scripts/ directory in the repository is written in Shell, which is the repository's declared primary language, but the README does not document which specific scripts are included there or what each one does. Teams that want to customize or extend the templates should explore the scripts/ directory directly after cloning.

Manual Feishu Whiteboard Styling as the Alternative

The direct alternative to beautiful-feishu-whiteboard is generating Feishu whiteboards with an agent that has no palette skill loaded. An agent without this skill either makes arbitrary color choices based on its own training or requires the engineer to specify colors inline for every whiteboard request. Neither approach captures the Feishu renderer's documented constraints.

A human designer using Feishu's native color picker can produce similarly styled whiteboards without this skill, but that approach does not scale to agent-generated content. The skill's value proposition applies specifically to automated whiteboard generation, where consistent palette application across many agent runs matters more than the flexibility to pick arbitrary colors. For one-off whiteboard creation by a designer, Feishu's built-in tools suffice.

For teams that want to see how each palette looks before committing, the README links to a Feishu doc version of the introduction at bytedance.us.larkoffice.com that includes usage-scenario showcases. Browsing that document lets a designer evaluate the 35 styles in context, which is more informative than reading the palette names from CATALOG.md alone.

Maintenance and MIT License

The last push to the repository was on 2026-06-11. The repository is not archived and has no GitHub releases. The primary language is Shell, and the templates directory holds the palette definitions.

The MIT license permits commercial use, modification, and redistribution. Teams embedding this skill in a proprietary agent workflow may do so without opening their own code. The lark-cli dependency carries its own license from ByteDance, and the Feishu and Lark platforms are commercial services; any team deploying this skill for automated whiteboard creation at scale should check the ByteDance developer terms for API usage limits. The repository has no GitHub releases, and the README lists no versioning scheme for the 35 templates or a migration path if templates change between commits.

Editorial conclusion

Teams that use Claude Code or other SKILL.md-compatible agents to produce Feishu or Lark whiteboards will benefit from this library because RULES.md encodes renderer constraints that are not documented by ByteDance. Designers who need a layout engine or a standalone whiteboard tool should look elsewhere; this skill gives agents palette and rendering rules, not layout composition logic. Verify that your agent runtime reads SKILL.md before adopting it. The last push was on 2026-06-11, under the MIT license.

Frequently asked questions

Official sources

1. Issues
2. License: MIT
3. README
4. zarazhangrui/beautiful-feishu-whiteboard on GitHub