AgentLint – Design Audit for AI Coding Agents

About this plugin

Designs that AI agents can build right the first time.AI coding agents like Codex, Claude Code, Cursor and GitHub Copilot build exactly what they can read from your file. Unnamed layers, hard-coded colors, detached components, absolute layouts and missing states all get guessed, and guesses become bugs.AgentLint audits your designs the way an agent reads them. Scan a selection, a page, or your entire file in seconds.See where you stand• AI Readiness Score (0–100) across 8 categories: naming, tokens, components, variables, responsive, states, accessibility and agent ambiguity• Agent Misinterpretation Risk: how likely an agent is to build the wrong thing, and exactly why• 48 checks, including unnamed layers, hard-coded values, missing variables, detached components, missing loading/empty/error/disabled states, contrast, touch targets, reading order and placeholder copyFix it fast• Recommendations ranked by how many points each one gains you• One-click safe auto-fix: semantic layer names, exact-match token and style binding, reading order. Every fix is visually neutral, re-verified before it's applied, and undoable in one step.• Before/after comparison and audit history to track progressHand off with confidence• Agent Contract: components, tokens, responsive rules, states, accessibility requirements and open questions, in one spec• Ready-made exports for Codex (AGENTS.md), Claude Code, Cursor rules, Copilot instructions, MCP, Markdown and JSON• PDF report of every issue, grouped by rule with fixes, ready to share with your teamPrivate by design. No network access, no account, no tracking. Your designs never leave Figma.Free: 5 audits a week (up to 3 frames each), readiness score and risk, your top issues and recommendations, and a PDF report.Pro: unlimited audits, complete-file scans, every issue with agent-impact analysis, auto-fix, the Agent Contract and all exports. Purchased and managed securely through Figma.