| name | figma-audit |
| description | Accepts a Figma frame URL, checks whether the structure is easy for AI to understand across 4 dimensions, and outputs an improvement report. |
| argument-hint | <figma-url> |
| allowed-tools | mcp__plugin_figma_figma__get_design_context, mcp__plugin_figma_figma__get_metadata |
Figma AI-Readiness Audit
Analyze the Figma frame URL provided in $ARGUMENTS and evaluate whether it is structured in a way that allows AI to implement it accurately.
Output language: Respond in the same language the user is using in this conversation.
Step 1: Parse the URL
Extract from $ARGUMENTS:
figma.com/design/:fileKey/...?node-id=X-Y → fileKey and nodeId (convert X-Y to X:Y)
Step 2: Fetch data
Call get_design_context and get_metadata in parallel:
get_design_context: retrieves generated code, screenshot, annotations, and style info
get_metadata: retrieves layer hierarchy and node ID structure
Parameters for get_design_context:
disableCodeConnect: true
clientLanguages: "unknown"
clientFrameworks: "unknown"
Step 3: Analyze across 4 dimensions
Dimension A: Layer naming
Good naming criteria:
- Meaningful names representing section or role (e.g.
Hero, CardBase, Footer)
- Elements inside components have role-based names (e.g.
HeroHeading, CTA, Button)
- No duplicate names among sibling elements
Problem patterns (checklist):
Dimension B: Auto-layout usage
Good state criteria:
- Auto-layout is applied to vertical and horizontal groups
- Spacing is managed via padding and gap (not manually positioned)
Problem patterns (checklist):
Dimension C: Variables and tokens
Good state criteria:
- Colors are defined as Figma variables and applied (generated code contains
var(--name))
- Spacing (margins, sizes) is managed with tokens
- Text styles are defined and applied (e.g.
heading/level_1)
Problem patterns (checklist):
Dimension D: Component and instance structure
Good state criteria:
- Repeatedly used elements are componentized
- Instances are used consistently (not manual copies)
Problem patterns (checklist):
Step 4: Output the report
Use the following format:
Figma AI-Readiness Audit Report
File: [file name]
Target node: [node name] (nodeId)
Screenshot: [display the screenshot]
Overall score: X / 20
| Dimension | Score | Rating |
|---|
| A. Layer naming | X / 5 | 🟢 Good / 🟡 Room for improvement / 🔴 Needs fixing |
| B. Auto-layout | X / 5 | 〃 |
| C. Variables & tokens | X / 5 | 〃 |
| D. Component structure | X / 5 | 〃 |
Scoring criteria (5 points each):
- 5: No issues
- 3–4: Minor issues (1–2 items)
- 1–2: Needs improvement (3+ items)
- 0: Almost entirely unaddressed
Strengths
(List notable practices in this file that others can learn from)
Improvement suggestions
For each issue:
[Location / target of the issue]
- Problem: What is wrong
- Impact: How it hinders AI code generation
- Suggestion: Specific action to take in Figma
Summary
(1–2 sentences on the overall impression and the highest-priority improvement)
Error handling
- If fileKey or nodeId cannot be extracted from the URL: ask the user to provide a valid Figma frame URL
- If
get_design_context or get_metadata returns an error: report the error and ask the user to check access permissions for the Figma file
- If a component URL is passed instead of a frame URL: inform the user and suggest trying with a frame-level URL