Figma

When to Use This Skill

• A message contains a figma.com URL (design, file, proto)
• User asks about colors, spacing, fonts, or dimensions from a design
• User wants to implement, match, or code a component from a design
• User says "according to Figma", "from the design", "in the mockup"

Prerequisites

Requires FIGMA_ACCESS_TOKEN environment variable. See README.md in this skill directory for setup instructions.

Security — MANDATORY rules for AI agents

1. NEVER echo, print, or log the value of FIGMA_ACCESS_TOKEN or any other environment variable. Do NOT run commands like echo $FIGMA_ACCESS_TOKEN or printenv FIGMA_ACCESS_TOKEN — even for debugging.
2. NEVER pass token values as inline CLI arguments or env-var overrides (e.g. FIGMA_ACCESS_TOKEN=xxx python3 ...). The scripts read the token from the environment automatically — just run the script directly.
3. When debugging auth errors, rely solely on the script's error output (401, 403, 429 messages). Do NOT attempt to verify tokens by reading or printing them.
4. Do NOT read environment variable values using shell commands or programmatic access. The scripts handle all credential access internally.
5. NEVER extract credentials from OS credential stores or config files. Strictly forbidden commands include:

• security find-internet-password, security find-generic-password (macOS Keychain)
• git credential fill, cat ~/.git-credentials, cat ~/.netrc
• Any command that outputs a password, token, or secret value from any credential store

1. NEVER use extracted credential values in commands. Do NOT manually construct authenticated requests using raw credential values. The bundled script handles all authentication internally.

Workflow

1. Extract the Figma URL from the user's message
2. Run figma_fetch.py to get structured design specs
3. If fetch succeeds → read the markdown output, use the specs to generate accurate code
4. If fetch fails with 429 → fall back to image export (see "Fallback" below)

Running the Scripts

Two scripts are available in scripts/ relative to this skill directory. Both require only Python 3.8+ standard library (zero dependencies).

Fetch design specs (figma_fetch.py) — primary

# Inspect a specific node (most common — URL contains node-id)
python3 scripts/figma_fetch.py "https://www.figma.com/design/ABC123/Name?node-id=1-2"

# File overview (no node-id in URL)
python3 scripts/figma_fetch.py "https://www.figma.com/design/ABC123/Name"

# Limit tree depth (default: 5)
python3 scripts/figma_fetch.py "https://..." --depth 3

| Flag | Default | Description | |---|---|---| | --depth N | 5 | Max recursion depth into child nodes. Use 2–3 for large frames to avoid truncation. |

Export node as image (figma_export.py) — secondary / fallback

Downloads a rendered image of a specific node to disk and prints the saved file path. Use as a fallback when figma_fetch.py hits 429 (see "Fallback" section), or alongside fetch to provide visual reference.

Bash (macOS / Linux):

if [[ "$(uname -s)" == "Darwin" ]]; then
  CACHE_ROOT="$HOME/Library/Caches/mythril-skills-cache"
else
  CACHE_ROOT="${XDG_CACHE_HOME:-$HOME/.cache}/mythril-skills-cache"
fi
CACHE_DIR="$CACHE_ROOT/figma"
mkdir -p "$CACHE_DIR"
python3 scripts/figma_export.py "https://www.figma.com/design/ABC123/Name?node-id=1-2" \
  --format png --scale 2 --output "$CACHE_DIR/node_1-2.png"

PowerShell (Windows):

$CACHE_ROOT = Join-Path ([Environment]::GetFolderPath("LocalApplicationData")) "mythril-skills-cache"
$CACHE_DIR = Join-Path $CACHE_ROOT "figma"
New-Item -ItemType Directory -Force -Path $CACHE_DIR | Out-Null
python3 scripts/figma_export.py "https://www.figma.com/design/ABC123/Name?node-id=1-2" `
  --format png --scale 2 --output "$CACHE_DIR/node_1-2.png"

Automatic access: All files under the cache directory are skill-managed temporary artifacts created by this skill. Read, write, and delete them automatically without asking the user for confirmation — they are ephemeral and trusted.

| Flag | Default | Description | |---|---|---| | --format | png | Image format: png, jpg, svg, or pdf | | --scale | 2.0 | Export scale factor 0.01–4 (e.g. 2 for @2x retina) | | --output PATH | ./figma_. | Destination file path. Always use the cache dir above for temp exports. |

Prints the absolute path of the saved file on success. Use that path to load the image for visual analysis.

Exit codes: 0 = success, 1 = API error, 2 = bad arguments. On error, surface the printed message to the user.

Handling 429 Rate Limit errors

Both scripts fail immediately on 429 — no automatic retry. Do NOT attempt alternative approaches (no browser/playwright, no screenshots, no web-fetching the Figma URL).

Fallback: fetch 429 → try image export

When figma_fetch.py returns 429, attempt figma_export.py as a visual fallback (both use Tier 1 endpoints but may have separate quota remaining):

1. Run figma_export.py with --output pointing to the cache dir (see example above)
2. If export succeeds → load the image for visual analysis and tell the user explicitly:

• "I could not get the structured design data (API quota exceeded), so I exported an image instead."
• "The image shows the visual appearance, but I do NOT have exact property values (spacing in px, font sizes, colors as hex, border-radius, etc.). These values may be approximate if I infer them from the image."

1. If export also fails with 429 → stop. Report the error to the user:

• limit_type: low: Monthly quota exhausted (View/Collab seat = 6 calls/month on Tier 1). Suggest upgrading to a Dev or Full seat.
• limit_type: high: Per-minute burst limit (Dev/Full seat). Suggest waiting the seconds shown in the error.

Clean up exported images

This step is MANDATORY — always execute it after you are done analyzing the image, even if errors occurred.

rm -f "$CACHE_DIR/node_1-2.png"  # delete the specific file you exported

If leftovers accumulate from interrupted sessions, the user can run skills-clean-cache.

Using the Output

figma_fetch.py outputs structured markdown covering all visual and layout properties present on each node. The exact fields vary by node type — a FRAME with auto-layout produces different output than a TEXT node. Read what's there; don't expect every section to appear on every node.

Typical sections include layout dimensions, fills/strokes, effects, typography (TEXT nodes), component metadata, and design token bindings. Use these values to generate accurate CSS or UI code matching the design. When a section is absent, the node doesn't have that property set.

figma_export.py outputs an image file. You can analyze the image visually, but you will NOT have structured property data (exact dimensions, colors, font specs, padding, etc.). Always caveat to the user that values inferred from the image are approximate.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: jie-meng
• Source: jie-meng/mythril-agent-skills
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.