Figma cli

Skill wohsj110/figma_cli/skills/figma-cli

Use when reading Figma files or nodes, exporting node images or assets, listing Figma comments, components, styles, or variables through the figma-cli command line, or setting up Figma token credentials. Token-based reads work headlessly — no Figma app or browser required.From its SKILL.md

Install

npx -y skills add wohsj110/figma_cli --skill figma-cli

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, 581 tokens by cl100k_base, as published. Nobody here has run it

Figma CLI Skill

Use this skill when an agent needs stable, reproducible reads from the Figma REST API through figma-cli.

CLI Setup

Before using Figma commands, check whether figma-cli is available with command -v figma-cli.
If missing, install with brew install --cask wohsj110/tap/figma-cli; without Homebrew, go install github.com/wohsj110/figma_cli/cmd/figma-cli@latest (requires Go ≥ 1.26).

Credential Setup

Verify access first: figma-cli me. If it prints the current Figma user, credentials are ready.
If not configured, ask the user for a Figma personal access token, then store it in the OS keyring:

printf %s "$FIGMA_TOKEN" | figma-cli init --token-stdin

For CI or one-off runs, FIGMA_TOKEN env takes precedence over the keyring.
Never print, log, or hardcode token values. The CLI never prints the token; keep it that way in your own output too.

Core Rules

Paste full Figma URLs directly — file keys and node ids resolve automatically. Never guess a node id; if you don't have one, ask the user or list metadata first.
Default output is concise text. Add --raw only when the original API JSON is genuinely needed.
GET responses are cached for 15 minutes. Use --no-cache when freshness matters, --cache-ttl 1h to extend.
Request large files deliberately: start with --depth 2 before any deep dump.
--verbose prints request diagnostics without token values — use it when debugging API errors.

Commands

figma-cli me                                                        # current user / credential check
figma-cli file get "https://www.figma.com/design/KEY/name?node-id=1-2"
figma-cli file get FILE_KEY --depth 2                               # bounded file summary
figma-cli node inspect FILE_KEY --node 1:2 [--raw]                  # node summary / raw API JSON
figma-cli image export FILE_KEY --node 1:2 --format png --scale 2   # export image URL
figma-cli image export FILE_KEY --node 1:2 --format png --out ./dir # download asset locally
figma-cli comments list FILE_KEY                                    # review context
figma-cli components list FILE_KEY
figma-cli styles list FILE_KEY
figma-cli variables list FILE_KEY                                   # design tokens

Boundaries

This CLI is read-only inspection and asset export by design — it never writes back to Figma.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.