Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
DeepSeek
Muse+22 moreClaude CodeCursorCodex CLIDeepSeekMuse
Dots
Manus
Grok
Grok Bot+18 agentsFigma AI-Readiness Audit
Scores a Figma frame on how well AI can turn it into code: layer naming, auto-layout, variables and tokens, and component structure. Returns a 20-point report with concrete fixes to make in Figma.
Free
Figma AI-Readiness Audit
Example session with this skill installed
/figma-audit https://www.figma.com/design/<fileKey>/tidy-up-claude?node-id=0-49
- Read your context and instructions
- Compiled the figma ai-readiness audit
Figma AI-Readiness Audit Report
File: tidy-up-claude
Target node: backup (0:49) — HILMA, a Scandinavian interior e-commerce home page (1440 × 3598)
Overall score: 15 / 20
| Dimension | Score | Rating |
|---|---|---|
| A. Layer naming | 3 / 5 | 🟡 Room for improvement |
| B. Auto-layout | 4 / 5 | 🟢 Good |
| C. Variables & tokens | 4 / 5 | 🟢 Good |
| D. Component structure | 4 / 5 | 🟢 Good |
Strengths
- Every color and spacing value in the generated code resolves to a variable (
var(--brand/primary),var(--space/80)), with no stray hex values. - 13 text styles are defined and applied (H1/Light, Body, Display/Price, and so on), so AI can map typography to a real type scale.
- Header, Footer, Product Card and Button are real components used as instances. Button variants are named on a Size × State axis (Default / Hover / Disabled).
- Sections carry semantic names (
hero,featured-products,brand-story,newsletter) that translate directly into component names.
Improvement suggestions
Root frame backup (0:49)
- Problem: The top-level frame is named
backup, which says nothing about the page. - Impact: AI uses the root name for the page component, so it generates
export default function Backup(). - Suggestion: Rename it to the page's role, for example
HomePageorTop.
Text node Read Our Story (0:90)
- Problem: The layer name is the raw text content, while every other text node has a role-based name.
Impact: If the copy changes, the layer name no longer describes the element. It also breaks the naming pattern AI relies on.
- Suggestion: Rename it to
story-cta-label.
Four sibling divider-wrapper frames (10:2, 10:3, 0:94, 10:4)
Problem: Four identical wrapper frames with height 0 sit between sections, all with the same name. Each one wraps a single line.
- Impact: AI emits four
h-0wrapper divs with the divider exported as an image, instead of a border or an<hr>. - Suggestion: Drop the wrappers and use a top stroke on each section, or make one
Dividercomponent.
space/17 spacing token (newsletter input and Button)
- Problem: The 17px padding is tokenized, but it is off the 4/8px grid used everywhere else.
- Impact: AI can't map it to a standard spacing scale and ends up writing an arbitrary value like
py-[17px]. - Suggestion: Use
space/16and let the fixed 52px height handle the vertical alignment.
Product Card instance sizing (14:3 vs 14:8 / 14:13)
- Problem: The first card uses Fill container, but the other two are fixed at
405.333px.
Impact: AI generates flex-1 for one card and w-[405.333px] for the other two, so the grid won't reflow at other widths.
- Suggestion: Set all three instances to Fill container.
Summary
This frame is already in good shape for AI. Variables, text styles and components are all in place. Fixing the root name and the Product Card sizing would remove the two things most likely to produce visibly wrong code.
Connects securely to your tools. The creator never sees your data.
About this skill
AI design-to-code output is only as good as the Figma file it reads. A frame full of "Frame 123" layers, manually positioned children, hardcoded hex values and copy-pasted cards produces code with the same problems: meaningless component names, absolute positioning, arbitrary values and duplicated markup. This skill checks a frame for exactly those problems before anyone starts implementing it.
What it does
- Scores the frame 0–5 on each of 4 dimensions (20 points total): layer naming, auto-layout usage, variables and tokens, and component and instance structure
- Flags auto-generated layer names, duplicate sibling names, generic names and text layers named after their own content
- Finds manually positioned children, empty spacer frames used in place of gap settings, and fixed spacing values with no tokens behind them
- Checks whether colors and spacing are bound to variables and whether text styles are defined and applied, based on the code Figma actually generates
- Spots repeated structures that were never made into components, and state variations built as separate frames instead of variants
- Lists what the file already does well, so the report isn't only a list of problems
How it works
- Takes a Figma frame URL
- Fetches the frame's generated code, screenshot and style information, along with its layer hierarchy, through the official Figma MCP server
- Checks the frame against a fixed checklist for each dimension
- Returns a scored report. For each issue, it names the exact layer, explains how it would hurt the generated code, and gives a specific fix to make in Figma.
What it is not
This is a read-only report. It never edits the Figma file and does not write any code. It tells you what to fix, and you make the changes in Figma, or with a separate cleanup skill if you have one.
Requirements
Runs inside an agent connected to the official Figma MCP server, the same connection used for Figma-to-code work. It uses only get_design_context and get_metadata. No local file access and no other network calls.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Frontend & Web Apps.
Requires a Figma frame URL that includes a node-id. Audits one frame per run, not a whole file at once. Needs read access to the file through the Figma MCP server. It is a report only: it suggests fixes but does not apply them in Figma.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
6 uses
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
- Accounts
Figma
Permissions
Allowed Hosts
This skill's only external interaction is through the Figma MCP server the user's agent already has connected, using the read-only tools get_design_context and get_metadata (see https://www.figma.com/mcp-catalog/). There is no local file read/write, no direct network or HTTP calls, and no terminal use. It never writes to the Figma file.