A SwiftUI skill pack for the two features too new to know by heart

A small skill pack aimed squarely at the newest surface area

This is a compact set of Agent Skills for SwiftUI development, purpose-built for the very newest parts of Apple's platform rather than general iOS guidance. It follows the open Agent Skills specification, which means it is not tied to one coding agent; it works in Cursor, Codex, Claude Code and any other harness that reads the same format, and its own README states plainly that it requires Xcode with the iOS 26 SDK.

That last requirement is the thing to notice first, because it tells you exactly what this project is for. A skill pack that targets general SwiftUI patterns would be useful to almost anyone writing iOS apps. A skill pack that explicitly requires the newest SDK and covers only two specific, current-generation surfaces is a narrower and more deliberate choice: it exists because two particular pieces of the platform are new enough, and detailed enough, that an agent working from general training knowledge is likely to get them wrong or miss recent changes entirely.

Two skills exist today, one covering the platform's newest visual material system and one covering layout for a foldable iPhone form factor, with the README stating more will be added as separate folders under the skills directory as the platform continues to move.

Liquid Glass is exactly the kind of feature that ages a model's knowledge

The first skill covers implementing and reviewing the newest visual material introduced in the current OS generation, described as covering native APIs, common patterns, and pitfalls.

This is worth taking seriously as a category of risk rather than a convenience. A large language model's knowledge of an SDK is frozen at whatever point its training data ended, and a visual material system introduced with a new OS release is precisely the kind of API surface that did not exist, or existed in an earlier and different form, before that cutoff. Asking an agent to write code against it from memory risks two distinct failure modes: inventing plausible-looking APIs that do not actually exist, or applying an outdated mental model of how the effect composes with existing view modifiers.

A skill document that lays out the current, correct APIs and specifically calls out known pitfalls is a targeted fix for exactly that risk, and it is a more honest way to solve the problem than hoping the agent's training data happened to cover it well enough. The word review appearing alongside implement in the skill's description matters too, since it means the skill is meant to be used both when writing new Liquid Glass code and when checking whether existing code has already gotten it wrong.

Designing for a device most guidance has never considered

The second skill addresses SwiftUI layout for a folding iPhone form factor, covering size-based layout, keeping content usable across the fold, hinge interactions, and support for a second external display accessory.

Foldable-device support is a layout problem most SwiftUI developers, and most SwiftUI guidance written before such a device existed, have simply never had to think through. A foldable phone changes shape while an app is running, which breaks any layout logic that assumes a fixed aspect ratio or a single continuous surface, and a hinge is a physical seam that content needs to actively avoid running through rather than merely tolerate. Handling a second, separate display attached as an accessory is a further wrinkle again, since it raises the question of what content, if any, should appear there independently of the main screen.

None of that is guidance a general-purpose SwiftUI skill written even a year earlier would contain, because the hardware and the APIs supporting it are that recent. Packaging it as its own dedicated skill, separate from ordinary responsive-layout advice, is the right structural choice, since fold-aware layout is a genuinely distinct discipline from simply supporting different screen sizes.

Installing the whole pack, or just the one skill you need

Installation offers the same two-tier choice this category of tool increasingly settles on. The recommended path uses a skill-installer command-line tool that copies skills into whichever agent's own directory it detects:

npx skills add FloWritesCode/fwc-swiftui-skills

The same tool takes a `--skill` flag to install a single named skill instead of the whole pack, which matters here specifically because the two skills solve genuinely unrelated problems: a developer working only on visual polish has no use for the foldable-layout skill, and a developer targeting only the folding form factor has no use for material-rendering guidance. Installing selectively rather than as a bundle keeps an agent's loaded context relevant to the actual work rather than padded with material that will never come up in that particular app.

For agents outside the installer's supported list, a plain clone-and-copy works just as well, since a skill is simply a directory with a properly formatted entry file, and any agent honouring the same open specification can use it once that directory sits in its expected skills location.

A repository built to make adding the next skill easy

The repository layout is minimal by design: a README, a licence file, and a skills directory holding one subdirectory per skill, each containing its entry file and an optional longer reference document for material too detailed to keep in the entry file itself.

The README's own contribution instructions are three steps: create the new skill's entry file with the required frontmatter, add a row to the README's table describing it, and keep the skill focused, pushing long reference material into the separate reference file rather than bloating the entry point. That is a genuinely low-friction process for a maintainer adding the next skill as Apple's platform keeps moving, and it explains why the project can credibly promise more skills will appear as separate folders over time rather than the two it ships with today.

Keeping each skill's entry file thin, with detail deferred to an optional reference document loaded only when needed, mirrors the same lazy-loading design that makes larger skill collections practical for an agent to hold in context, and applying that discipline even to a two-skill pack suggests the structure was chosen with future growth already in mind rather than retrofitted later.

What to keep in mind before relying on it

The most important thing to understand about a project like this is that its value is tied entirely to how current it stays. A skill covering a brand-new visual material system or a very recent device form factor is valuable precisely because it is current, and that value decays the moment Apple ships a change the skill has not yet caught up with. The README gives no explicit statement of an update cadence, so a prospective user should judge currency from the repository's own activity rather than assume it is guaranteed.

On that count the signal is reasonably good: the project is MIT licensed, with the last push on 2026-09-10, and reports 317 stars, 15 forks and no open issues, so it is a small but apparently active and well-regarded project rather than an abandoned one.

Before adopting it, three steps in order. Confirm your Xcode installation actually has the iOS 26 SDK, since the README states this as a hard requirement rather than a recommendation. Install only the specific skill matching the work in front of you, using the installer's skill flag, rather than the whole pack, since the two current skills solve unrelated problems. And periodically check the repository for new skills or updates to the existing ones, since a project built around the newest parts of a fast-moving platform is only as useful as how recently it was last touched.

Editorial conclusion

This pack fits an iOS developer working against the current iOS 26 SDK who needs an agent to get Liquid Glass or foldable-device layout right, precisely because those two surfaces are recent enough that a model's own training data is the least reliable source for them. Install only the skill matching your current work using the installer's skill flag, since the two skills address unrelated problems and there is little reason to load both into an agent's context at once. Confirm your Xcode installation actually carries the iOS 26 SDK before relying on either skill, and treat the whole project's value as tied to how recently it was updated, checking back for new skills or revisions as Apple's platform continues to move.

Frequently asked questions

Official sources

1. FloWritesCode/fwc-swiftui-skills on GitHub
2. Issues
3. License: MIT
4. README