UX designer CV skills to list after Figma agents
Quick answer: A UX designer CV in late 2026 needs three things the pre-agent version didn't: named agent and code tooling (Figma Specs, code layers, Dev Mode, MCP, design tokens), bullets showing you directed and reviewed agent output rather than just prompted it, and outcome numbers like task completion rate or shipped release count. Cut tool logos, "prompt engineering" and anything the portfolio proves better.
Is your CV good enough?
Upload your CV and get an instant AI score out of 100, an ATS-compatibility rating and a breakdown across five categories — free.
What changed when Figma shipped agents and code layers?
Figma made agents an official part of design work on October 14, 2026. At its first Config event in India, the company launched Specs — a Markdown editor where you write your design language once so that teammates and agents read the same rules — and moved code layers from private beta into open beta on every plan. The Figma Agent is now generally available, and the platform added twenty new MCP integrations, among them OpenCode, Dia, Junie by JetBrains and Meta Muse. Code Sources, still in closed beta, lets a team point agents at a GitHub repo or an npm package so generated components match what engineering actually ships. This isn't a side panel. It's the canvas.
Code layers are the part that reshapes the job. You can now build, edit and compare code-backed prototypes on the design canvas itself — prompting the Figma agent, importing from a repo, capturing a live site, or wiring in a coding agent over MCP — and convert existing designs into something you can genuinely test for behaviour, accessibility and consistency. Figma's own chief design officer put a number on the pressure that creates: code output has grown seventeenfold while shipped releases grew just 1.3 times. Exploration got cheap. Judgement about what deserves to ship got expensive, and that gap is exactly what hiring managers are now screening for.
So the CV has to change with it. A skills section that reads "Figma, Sketch, Adobe XD, prototyping, wireframing" describes a 2021 workflow and tells a recruiter nothing about whether you can brief an agent, review what it returns, and defend a design system against a flood of generated components. Job ads for product designers already name Claude Code, Cursor, Code Connect, design tokens, Figma variables and React in the same breath as user research and accessibility. Keyword matching is literal — if the terms aren't in your document as plain nouns, nothing matches them for you.
Exploration volume has outrun shipped work, which is why judgement is now the thing designers are screened for.
Which UX designer CV skills do recruiters and ATS match on now?
The highest-weight terms on a product designer CV in 2026 are Figma plus its newer surfaces (Specs, code layers, Dev Mode, variables, Auto Layout), design systems vocabulary (tokens, components, variants, Code Connect), user research and usability testing, information architecture, accessibility to WCAG 2.2, and at least one front-end term you can actually back up. Add MCP if you've connected a coding agent to design context, because that acronym now appears in job ads as a discrete requirement. Aim for roughly 15 to 25 real skills: eight to twelve hard skills, three to five named tools, a couple of methodologies, and the research platforms you've used.
Build that list from evidence, not instinct. Pull five to ten live postings for the exact role you want, highlight every term that repeats across three or more of them, and keep only the ones you could survive a follow-up question about. Treat each one as a plain noun — parsers match "usability testing", not "testing usability with real humans because empathy matters". Where a posting uses a specific phrasing, mirror that phrasing when it's honestly true of your work. This is dull, mechanical work, and it's the single highest-return hour you'll spend on your CV.
The category most designers skip is metrics. Task completion rate, error rate, time-on-task, SUS, CSAT, conversion and retention are all searchable terms, and they're the vocabulary that lets a hiring manager read your bullets as outcomes rather than activity. If you'd rather not assemble and format this by hand, you can build the skills section against one specific job ad and let the template handle structure while you supply the substance. Just don't let the tool decide what's true about you.
Pre-agent lines on the left, what 2026 product design postings actually match on the right.
| Pre-agent skills line | What a 2026 posting matches on |
|---|---|
| Figma, Sketch, Adobe XD | Figma, Specs, code layers, Dev Mode, variables, MCP |
| Prototyping | Code-backed prototypes, React/TypeScript prototypes, interactive flows |
| Design systems | Design tokens, component architecture, Code Connect, Style Dictionary |
| Wireframing, user flows | Agentic and conversational interaction patterns, multi-step task flows |
| Prompt engineering | Spec writing, agent evaluation, design QA of generated output |
| Passion for pixel-perfect UI | WCAG 2.2 conformance, task completion rate, time-on-task |
How do you word a bullet about directing agents without sounding like a prompt jockey?
Write the bullet around the decision you made, not the prompt you typed. "Prompted AI to generate 40 screen variants" reads as button-pressing; "Wrote the design spec that three coding agents built against, cutting handoff questions from 20 per sprint to 4" reads as leadership. The verb matters: directed, specified, reviewed, rejected, constrained, standardised. Agents produce volume, so your value sits in the constraints you set up front and the quality bar you enforce afterwards. Any bullet where the agent is the subject and you're the operator undersells you. Flip it so you're the subject and the agent is the instrument.
Three shapes work reliably. First, the specification bullet: "Defined component rules in a shared spec file so generated UI matched the design system without manual rework." Second, the review bullet: "Audited agent-generated flows for accessibility, catching 31 WCAG 2.2 failures before release." Third, the throughput bullet: "Shipped 6 code-backed prototypes in two weeks, replacing static mockups in usability tests and raising task completion from 61% to 84%." Each one names a method, a number and a consequence. That's what survives both a parser and a design director reading forty CVs on a Friday afternoon.
Be straightforward about how much of the work was yours. Designers worry that mentioning agents devalues their craft, and the opposite is true in 2026 — hiring teams assume you use them and want to know how well. The same logic applies to your case studies, where how to disclose AI use at the process level matters more than a blanket label. One line in your skills section listing the agents and coding tools you work with, then evidence in the bullets, beats sprinkling "AI-powered" across every role.
Flip the subject: you direct, specify and reject — the agent executes.
Is your CV good enough?
Upload your CV and get an instant AI score out of 100, an ATS-compatibility rating and a breakdown across five categories — free.
Where should design-to-code work go on a product designer CV?
Put design-to-code work inside your experience bullets, not in a separate "technical skills" annex at the bottom. Recruiters read top-down and stop early, and a React mention stranded below your education does nothing for a role that asks for designers who prototype in the real codebase. Name the stack where you used it: "Built the onboarding flow as a code layer in React and TypeScript, handed to engineering as production-ready components." Principal and founding roles now explicitly ask for shipped front-end code, and some ask for recent production commits. If that's you, it belongs in your first three bullets, not your last.
Be precise about the altitude you work at, because overclaiming here is easy to catch. There's a real difference between writing throwaway prototype code, contributing tokens and components to a shared library, and merging production front-end work. Say which one you do. "Contributed 14 components with tokens and documentation to a design system used across 5 products" is a credible, verifiable claim. "Full-stack designer" is not, and it invites a technical screen you'll fail. One concrete, modest, checkable claim converts better than a broad one, and it's the kind of detail an interviewer builds a conversation around.
Keep the portfolio and the CV doing different jobs. The CV gets you found; the portfolio gets you chosen. Screening software can't open your case studies, so every skill you demonstrate visually must also appear as text in the document, and your portfolio link should sit in the contact header as plain text because hyperlinks get stripped. Referencing a specific case study inside a bullet — "see the agentic search case study" — does double duty: it adds matchable terms and pre-sells the click to a human reader.
CV text has to restate what your portfolio shows, because parsers never open the link.
What should you cut from a pre-agent skills section?
Cut dead tooling, cut filler adjectives, and cut anything your portfolio already proves better. Sketch and Adobe XD now signal that your last tool decision was made years ago — list them only if the posting does. "Prompt engineering" as a standalone skill has aged badly in under two years; the durable version is spec writing and agent evaluation, which is what Figma's Specs release just institutionalised. Drop "pixel-perfect", "passion for beautiful experiences" and "user-centric mindset", because there's nothing in them to match and nothing to verify. Also cut skill-rating bars: they consume space, parse as nothing, and quietly cap you at whatever percentage you drew.
Trim the archaeology too. A junior CV listing six university projects and a senior CV listing a 2014 agency internship both bury the work that matters. Give your last two roles the detail and compress everything older into one line each. Certifications still carry weight as searchable terms — the Google UX Design Certificate, NN/g, IxDF — but they sit under Education, not alongside your core skills, and they don't substitute for shipped work. If a line doesn't name a tool, a method, a metric or a credential, it's taking up room that a real bullet could use.
Then there's format, which quietly kills more designer CVs than content does. Two-column layouts, icon rows, text boxes and tables are exactly what designers reach for and exactly what breaks extraction, so the way a designed CV scans versus how it looks is worth checking before you apply anywhere. Standard headings — Experience, Skills, Education, Portfolio — parse cleanly and still look sharp in a single column. Running your file through a free CV check that scores ATS compatibility and tech stack takes a minute and tells you whether your new agent keywords actually made it out of the PDF.
Designers break their own CVs with two-column layouts more often than with weak content.
Frequently asked questions
Do I need to write production code to get a product designer job now?
Not for most roles, but you do need to work credibly near code. Mid-level postings ask for code-backed prototypes, design tokens and design-system contribution. Principal, founding and AI-product roles increasingly require shipped front-end work in React or TypeScript. Be specific on your CV about which altitude you operate at — prototype code, library contributions, or merged production work — because overstating it collapses in the first technical screen.
How many AI and agent tools should I list on my CV?
Three to five, and only ones you've genuinely worked in. A long list of agent names reads as padding and invites questions you can't answer. Name the platform surfaces you actually use — Figma code layers, Specs, Dev Mode, an MCP connection, whichever coding agent your team runs — then prove each one in a bullet with a method and a number. Depth on three beats a logo wall of twelve.
My last employer banned AI tools. Am I out of the running?
No, but don't leave the gap unaddressed. Write the transferable half honestly: design-system governance, written specs, component architecture, accessibility auditing and design QA are the exact skills agent workflows depend on. Then build one personal or side project using code layers and an agent, and write it up with an outcome. A single concrete example beats a skills line claiming familiarity you can't demonstrate in conversation.
Should my portfolio link go on the CV if screening software can't open it?
Yes — put it in your contact header as plain text, not a hyperlink, because links are commonly stripped during extraction. Human reviewers reach your portfolio after the automated screen, so it has to be there for them. The important part is redundancy: any skill, method or tool your case studies demonstrate must also appear as text in the CV, since only the text gets matched against the job description.
Is your CV good enough?
Upload your CV and get an instant AI score out of 100, an ATS-compatibility rating and a breakdown across five categories — free.