Boost Feishu CLI with Claude Code: Create Diagrams and PPTs in One Command

Feishu CLI and Whiteboard Skill Overview

Feishu CLI is an open‑source (MIT) tool that exposes 18 Feishu business domains via 200+ commands and 26 AI Agent Skills. The lark-whiteboard skill enables reading, creating and editing whiteboard objects directly in Feishu documents as vector nodes rather than static images.

Installation

Prerequisite: Node.js with npm/npx (example: node -v → v24.13.0, npm -v → 11.6.2).

Install with a single command: npx @larksuite/cli@latest install The installer asks “Ok to proceed?” (answer y), language selection, then globally installs @larksuite/cli (v1.0.39) and the 26 skills under ~/.claude/skills/, including lark-whiteboard.

Verify installation:

lark-cli --version
lark-cli version 1.0.39

During installation a QR‑code is shown; scanning it links the machine to a Feishu account and creates an app credential. After linking commands such as lark-cli docs and lark-cli whiteboard become usable.

Whiteboard Skill Capabilities

Read a document, locate embedded whiteboards, and export previews as SVG, Mermaid, PlantUML, or raw JSON.

Provide an empty board together with SVG/Mermaid/PlantUML/DSL input; the skill renders an editable board object back into the document.

Edit an existing board – change text, colors, or structure by double‑clicking a node.

Example Workflow: Generating an Editable PPT from a PDF

Command:

lark-cli design-lark-chart Claude-for-the-financial-industry-05012026_v4.pdf --style anthropic --ratio 16:9

The CLI reads the PDF, creates a 16:9 information‑rich PPT, inserts it into a new Feishu document, and the resulting slides are editable vector boards.

Demonstrated Use Cases

Mind map – supply a URL, the skill extracts core content and produces a modifiable mind map.

Cheat sheet – parse the larksuite/cli README, select appropriate chart types, and render a three‑column cheat sheet with code snippets.

Sankey diagram – feed local data, transform it into a flow diagram (e.g., e‑commerce traffic from source → landing page → user actions).

UI mockups – generate desktop and mobile data‑analysis dashboards from extracted fields.

Product roadmap – draw a timeline of Claude model releases without manual design.

Swimlane diagram – create an e‑commerce order process diagram (5 lanes × 6 stages, 9 connectors).

Architecture diagram – visualise a five‑layer micro‑service stack (access → gateway → service → middleware → storage).

Illustrations – produce artwork, floor plans, and a periodic table, all as editable whiteboard objects.

Comparison with Traditional AI Drawing Tools

Output format : traditional tools output PNG/JPG images; Feishu whiteboard returns vector node objects.

Re‑editing : traditional workflow requires a new prompt and full regeneration; Feishu allows direct node editing (double‑click).

Collaboration : traditional approach uses screenshots → chat → comments; Feishu enables multiple users to edit the board simultaneously inside the document.

Knowledge retention : images are scattered on disk or in chat logs; Feishu stores the board with the document in the knowledge base.

Single‑point change : traditional method redraws the entire image; Feishu modifies only the affected node.

Target Audience

Product managers – create PRD flowcharts, swimlanes, state machines without leaving Feishu.

Backend engineers / architects – generate architecture, sequence and dependency diagrams that stay in sync with documentation.

Startup CEOs – produce business model canvases, value pyramids, growth‑flywheel diagrams that remain interactive for investors.

Technical bloggers – enrich articles with AI‑generated illustrations that can be exported directly.

Key Repository

Source code and releases are hosted at https://github.com/larksuite/cli