Week Two of Redacted: I Left Figma for Sketch

Notes from a week that was supposed to be about the game and turned into a workshop decision I couldn’t put off.

The past week on Redacted has mostly been a tools decision I didn’t expect to make: I moved my entire game UI from Figma to Sketch, manually, artboard by artboard, because I hit the free-tier ceiling on Figma and flinched at what I’d have to pay to keep going.

That framing understates what actually happened. What actually happened is that I’m now selecting design tools by a criterion I hadn’t consciously articulated a month ago: how well does this tool talk to the agent I’m building alongside?

That’s a shift. It’s worth naming out loud.

Hands-on with agentic support, not agentic instead of hands-on

The best modern design tool, as far as I can tell right now, is not a fully agentic one. It’s a hands-on tool that’s been made agentically readable, with a first-class agent surface layered on top.

The current model is Framer. They kept every hands-on tool intact — every nudge, every click, every direct-manipulation control that lets a designer move faster than any explanation could — and then added an agent tab, plus a local MCP that lets you bring whatever agent and budget you have to the tool. Not their agent. Your agent. That architecture matters more than any specific feature they’ve shipped, because it’s the acknowledgment that a working designer needs both kinds of leverage. Sometimes it’s faster to just click a color swatch than to describe the change I want. Sometimes it’s faster to hand off ten repetitive screens to an agent while I go do something else. The workflow that wins is both — and neither side is going away.

Rive is trying to reach the same shape from a different starting point. Its MCP tooling isn’t as clean off the shelf yet — more designer effort on-screen to get the agent to actually see what’s happening — but the direction is right. The catch is what each tool actually outputs. Framer is production-ready for websites, but doesn’t do apps. Rive makes gorgeous vector UI that you can’t drop straight into most production pipelines. Both matter. Both are still meaningful gaps in what an independent consultant like me can actually ship without an engineering team behind me.

Which brings me to Figma versus Sketch

I hit the Figma free-tier ceiling this week, and something in me didn’t want to pay for it. I already pay Adobe more than I want to. And at the end of the day, Figma is a web app that has been racing to add production-adjacent tools — Figma Sites, Figma Make, and so on — that I don’t rate. Framer beats Figma at websites. Figma Make gets beaten by Lovable, v0, Base44, and Claude Designer for actual output and code management. It’s a fine tool. It isn’t one I want a monthly bill for anymore.

Then a YouTube video surfaced in my feed comparing MCP-driven output between Figma and Sketch, and Sketch won the head-to-head. Which prompted the question I hadn’t asked in six years: wait, is Sketch still around?

It is. It’s macOS-native, which means it feels the way a serious design tool should feel — the kind of polish I forgot to miss in the years I spent inside browser tabs. It offers a permanent local license, which no one else in this space is still selling. It has bring-your-own-agent MCP support built in. And it’s a few dollars a month cheaper than the Figma tier I would have needed.

I switched.

About that “.fig import” feature

Sketch has an import option for .fig files, and if you’re considering the migration, I want to be straight: it does not save you from a manual pass. Basic typography and image frames survive. Anything more complicated — components, auto-layout, most interactive states, most nested groups — either doesn’t render, doesn’t come across at all, or comes across so scrambled that rebuilding from scratch is faster than repairing. I scrapped most of what came through and rebuilt view by view.

That’s not a knock on Sketch. It’s what an interop feature between two proprietary formats is going to be until one buys the other. Plan for a manual migration.

While I’m on the subject of design tools I have opinions about

Three quick notes on tools I’ve spent time in and have things to say about.

Pencil. I was forced to use Pencil for three months at the end of my last full-time role. Under conditions I couldn’t renegotiate, it cut my design throughput by an estimated 70% and my output quality by roughly 50%. The specific reasons: it doesn’t save to the cloud, so client review requires either a live-call demo or building your own HTML preview CI/CD pipeline; copying assets between documents to share a library corrupts the destination file and produces duplicate library references; library updates don’t propagate downstream automatically; and agent context is blown constantly by the tool’s screenshot-based visibility model. I built custom prompts and skills specifically to repair the tool’s schema corruption on the fly. That is the wrong workflow. That’s the workflow where the tool is fighting the designer.

Paper.design. I tried it once after I went independent. The one legitimately interesting thing about it is that the canvas renders actual React rather than a proprietary JSON, which enables some genuine copy-paste round-tripping. Beyond that, I didn’t stay long enough to form a full opinion. I need to enjoy my tools, and I didn’t enjoy this one enough to keep going.

Subframe. I respect what Subframe is doing. It’s not the Figma killer it was marketed as, but it’s genuinely useful in a narrow, high-value niche. I use it. I don’t have a bigger claim to make about it than that.

The next move for Redacted

The reason all of this matters for the game is that Sketch, if I can make it a reliable source of truth for Rive, becomes my rough-comp environment for Redacted’s real UI. I don’t have the muscle memory in Rive yet — it’s at least twice as complex a UI as Sketch even for simple design work — and building comps in Sketch first is faster while I’m learning Rive properly.

The blocker is that there’s no off-the-shelf Sketch-to-Rive pipeline. So this week’s work is going to be a custom skill or two — small elements at a time, screenshots and properties as reference material — to get the migration reliable enough to build on. When those skills exist and are trustworthy, I’ll put them on GitHub. That’s likely a post of its own once there’s something honest to say.

This didn’t look like game work from the outside. From the inside, all of it was. Migrating by hand meant real design decisions along the way — not re-tracing what existed but tightening what wasn’t right yet — and the UI is closer to semi-final than it was a week ago. The Sketch-to-Rive skill work now has a real source of truth to work from. Some weeks the workshop decision and the actual work turn out to be the same work.