Optimizing Lazy Layouts logo

Optimizing Lazy Layouts

Community
skydoves
optimizing-lazy-layouts

Use this skill to fix scroll jank, lost item state, and broken animateItem() animations in LazyColumn, LazyRow, LazyVerticalGrid, and LazyHorizontalGrid. Covers stable item keys, contentType for mixed-type feeds, Modifier.animateItem() requirements, hoisting modifier chains and painters out of the items lambda, and validating item composable stability. Use when the developer mentions LazyColumn jank, dropped frames while scrolling, items losing scroll state on insert/remove/reorder, mixed feeds of cards/headers/ads feeling sluggish, animateItem() not animating, RecyclerView view-type analog, key parameter, or contentType parameter. The prefetch-window tuning lives in a sibling skill.

Overview

Publisherskydoves
Repositorycompose-performance-skills
Skill nameoptimizing-lazy-layouts
Stars
515
Forks
19
Bundled files
Instructions only
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by skydoves on GitHub. Read the source before you install it.

Installation

Install the Optimizing Lazy Layouts AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/skydoves/compose-performance-skills/tree/main/lists/optimizing-lazy-layouts
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/skydoves/compose-performance-skills.git /tmp/compose-performance-skills
mkdir -p .claude/skills
cp -r /tmp/compose-performance-skills/lists/optimizing-lazy-layouts .claude/skills/optimizing-lazy-layouts
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Optimizing Lazy Layouts in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Optimizing Lazy Layouts on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Optimizing Lazy Layouts is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Optimizing Lazy Layouts — Keys, contentType, and animateItem()

Lazy layouts compose only what's visible, but two things still cost: re-composition of items that should have been reused (missing key), and per-item allocation that compounds with scroll velocity (missing contentType, modifier chains created inside items { }). Both have a one-line fix. This skill teaches Claude how to apply that fix correctly and to validate that item composables are themselves skippable. Prefetch tuning is a separate concern — see ../configuring-lazy-prefetch/SKILL.md.

When to use this skill

  • The developer reports scroll jank, dropped frames, or stutter on a LazyColumn, LazyRow, LazyVerticalGrid, or LazyHorizontalGrid.
  • Items lose scroll position, focus, or composition state on insert, remove, or reorder.
  • A mixed-type feed (cards + headers + ads + carousels) feels sluggish even though each individual row is lightweight.
  • Modifier.animateItem() was added but no animation runs on inserts or removals.
  • The compiler report shows item composables as unstable/non-skippable, or @TraceRecomposition shows item composables recomposing on every scroll tick.

When NOT to use this skill

  • The bottleneck is the prefetch window (heavy items, high-velocity scroll, want a wider ahead/behind window) → use ../configuring-lazy-prefetch/SKILL.md.
  • The item composable itself takes an unstable parameter (List<Foo>, Flow<Foo>, a domain var) → first run ../../stability/diagnosing-compose-stability/SKILL.md and then ../../stability/stabilizing-compose-types/SKILL.md.
  • An animation inside an item reads state.value in Composition phase, recomposing the row every frame → use ../../recomposition/deferring-state-reads/SKILL.md.
  • Scroll position derivation (e.g. firstVisibleItemIndex == 0) is the hot path → use ../../recomposition/choosing-derivedstateof/SKILL.md.

Prerequisites

  • Compose Foundation 1.7+ for Modifier.animateItem() (the GA replacement for the experimental animateItemPlacement).
  • Kotlin 2.0.0+ with org.jetbrains.kotlin.plugin.compose applied. Strong Skipping is on by default; non-skippable item composables become amplified at scroll speed.
  • A real device + release build for measurement. Skydoves hot take #5: debug builds lie (Live Literals, interpreted mode). See ../../measurement/generating-baseline-profiles/SKILL.md when ready to measure.

Workflow

  • 1. Audit every items(...) call. Walk every LazyListScope.items(list), items(count), itemsIndexed(list), and the LazyGridScope equivalents. For each, decide: does each element have a stable identity that outlives a single composition? If yes — and it almost always does — supply key = { it.id } using a server-side stable ID. MUST NOT use the list index, UUID.randomUUID() evaluated per emission, or hashCode() of a mutable object.
kotlin
// WRONG
LazyColumn { items(snacks) { snack -> SnackRow(snack) } }
// WRONG because: index-based identity → insert/remove discards composition state and breaks animateItem().
kotlin
// RIGHT
LazyColumn {
    items(
        items = snacks,
        key = { it.id },
        contentType = { it::class },
    ) { snack ->
        SnackRow(snack, Modifier.animateItem())
    }
}
  • 2. Add contentType for heterogeneous lists. Lazy layouts maintain a per-type composition cache analogous to RecyclerView's view-type. When item N + 1 has the same contentType as a recycled slot, the cached composition is reused; otherwise it is discarded and rebuilt. For homogeneous lists Compose infers a single content type and contentType is optional. For mixed feeds (cards, headers, ads, carousels, dividers) MUST supply a stable type discriminator.

  • 3. Validate item composable stability. Run ../../stability/diagnosing-compose-stability/SKILL.md. If the item composable accepts an unstable parameter, no amount of key/contentType work will help — the row recomposes on every scroll-driven snapshot tick anyway. Fix with ../../stability/stabilizing-compose-types/SKILL.md before tuning further.

  • 4. Hoist allocation-heavy values out of the items lambda. The items lambda runs once per item per scroll-driven (re)composition. Painters, color resolutions, shapes, and BorderStroke instances built inside the lambda are reallocated each pass. Hoist constants and remember-based caches above the LazyColumn or to the call site. Modifier chains are themselves cheap because Compose deduplicates them structurally — hoist a Modifier only when profiling proves it matters.

  • 5. Add Modifier.animateItem() for visual continuity. Pair with a stable key. The animation runs on inserts, removals, and reorders; without key the animation cannot bind to identity and silently no-ops. The default fade-in / fade-out / placement spring is usually correct; tune with fadeInSpec, fadeOutSpec, placementSpec only when the design system requires it.

  • 6. Cache common painters / colors / shapes outside the items block. painterResource(...), MaterialTheme.colorScheme.surface, RoundedCornerShape(...) resolutions on every item composition add up. Hoist to the screen-level composable and pass down, or remember once at the LazyColumn parent.

  • 7. Verify with @TraceRecomposition and Layout Inspector. During a controlled scroll, expect each item composable to recompose at most once per real state change — not per scroll tick. Layout Inspector → Recomposition Counts column should plateau, not climb monotonically.

Patterns

Pattern: missing key

kotlin
// WRONG
LazyColumn {
    items(snacks) { snack -> SnackRow(snack) }
}
// WRONG because: items default to index-based identity. On insert/remove/reorder, every position past the change point has a different "identity", composition state and scroll-restoration are lost, and Modifier.animateItem() has nothing to animate from.
kotlin
// RIGHT
LazyColumn {
    items(snacks, key = { it.id }) { snack -> SnackRow(snack) }
}

Pattern: random or unstable key

kotlin
// WRONG
items(snacks, key = { UUID.randomUUID() }) { snack -> SnackRow(snack) }
// WRONG because: a fresh key on every recomposition guarantees the cached composition is discarded every time — strictly worse than no key.
kotlin
// WRONG
items(snacks, key = { it.hashCode() }) { snack -> SnackRow(snack) }
// WRONG because: hashCode() of a mutable type changes when fields mutate, breaking identity continuity for the same logical item.
kotlin
// RIGHT
items(snacks, key = { it.id }) { snack -> SnackRow(snack) }

Pattern: mixed feed without contentType

kotlin
// WRONG
items(feed, key = { it.id }) { item ->
    when (item) {
        is FeedItem.Card -> CardRow(item)
        is FeedItem.Ad -> AdRow(item)
        is FeedItem.Header -> HeaderRow(item)
    }
}
// WRONG because: cached compositions of one type are discarded when scrolled into a different type's slot — every row crossing a type boundary is a fresh build instead of a recycled update.
kotlin
// RIGHT
items(
    items = feed,
    key = { it.id },
    contentType = { it::class },
) { item ->
    when (item) {
        is FeedItem.Card -> CardRow(item)
        is FeedItem.Ad -> AdRow(item)
        is FeedItem.Header -> HeaderRow(item)
    }
}

Pattern: Modifier.animateItem() without a stable key

kotlin
// WRONG
items(snacks) { snack ->
    SnackRow(snack, Modifier.animateItem())
}
// WRONG because: animateItem() binds animation state to the item's key. With no key, identity is index-based, so an insert at position 0 looks like every-row-changed and nothing animates correctly.
kotlin
// RIGHT
items(snacks, key = { it.id }) { snack ->
    SnackRow(snack, Modifier.animateItem())
}

Pattern: allocation inside the items lambda

kotlin
// WRONG
items(snacks, key = { it.id }) { snack ->
    val placeholder = painterResource(R.drawable.snack_placeholder)
    val border = BorderStroke(1.dp, MaterialTheme.colorScheme.outline)
    Card(border = border) {
        AsyncImage(snack.imageUrl, placeholder = placeholder)
    }
}
// WRONG because: painterResource resolution and BorderStroke allocation happen on every item composition; at high scroll velocity these compound into measurable allocation pressure.
kotlin
// RIGHT
@Composable
fun SnackList(snacks: ImmutableList<Snack>) {
    val placeholder = painterResource(R.drawable.snack_placeholder)
    val border = BorderStroke(1.dp, MaterialTheme.colorScheme.outline)
    LazyColumn {
        items(snacks, key = { it.id }, contentType = { it::class }) { snack ->
            Card(border = border) {
                AsyncImage(snack.imageUrl, placeholder = placeholder)
            }
        }
    }
}

Note: Compose deduplicates structurally-equal Modifier chains internally, so reallocating Modifier.fillMaxWidth().padding(16.dp) per item is a micro-optimization. Hoist a Modifier only when profiling identifies it as the bottleneck — premature remember { Modifier.… } adds noise without measurable benefit.

Pattern: unstable item composable swallows all gains

kotlin
// WRONG
@Composable
fun SnackRow(snack: Snack, tags: List<String>) { /* ... */ }

// Caller:
items(snacks, key = { it.id }) { snack ->
    SnackRow(snack, tags = snack.tags)
}
// WRONG because: List<String> is an unstable parameter under inference; every scroll-driven recomposition recomposes the row body even though the snack didn't change.
kotlin
// RIGHT
@Immutable
data class Snack(val id: Long, val name: String, val tags: ImmutableList<String>)

@Composable
fun SnackRow(snack: Snack) { /* ... */ }

items(snacks, key = { it.id }, contentType = { it::class }) { snack ->
    SnackRow(snack)
}

Cross-reference: ../../stability/stabilizing-compose-types/SKILL.md.

Pattern: LazyVerticalGrid with mixed spans

kotlin
// RIGHT — keys + contentType apply to grids identically
LazyVerticalGrid(columns = GridCells.Fixed(2)) {
    items(
        items = feed,
        key = { it.id },
        contentType = { it::class },
        span = { item -> if (item is FeedItem.Header) GridItemSpan(maxLineSpan) else GridItemSpan(1) },
    ) { item ->
        when (item) {
            is FeedItem.Header -> HeaderRow(item, Modifier.animateItem())
            is FeedItem.Card -> CardCell(item, Modifier.animateItem())
        }
    }
}

Mandatory rules

  • MUST specify a key for every items(...) block where item identity outlives a single composition (effectively: every list backed by domain objects).
  • MUST use server-side stable IDs as keys. MUST NOT use the list index, MUST NOT use UUID.randomUUID() evaluated per emission, MUST NOT use hashCode() of a mutable object.
  • MUST specify contentType for heterogeneous lists (cards + headers + ads, etc.). Use a stable type discriminator such as it::class or a sealed enum.
  • MUST NOT use Modifier.animateItem() without a stable key — the animation silently no-ops.
  • MUST validate item composable stability with ../../stability/diagnosing-compose-stability/SKILL.md before blaming the lazy layout. An unstable item parameter cancels every gain from key/contentType.
  • MUST NOT wrap items { } in extra inline composable wrappers (Row { items { } }) hoping to "force" skippability — Row/Column/Box are NOT restartable/skippable to begin with (skydoves hot take #3).
  • PREFERRED: combine with ../configuring-lazy-prefetch/SKILL.md for high-velocity scroll surfaces only after item-level fixes are in place.
  • PREFERRED: measure in release + R8 + on a real device (skydoves hot take #5) before declaring a fix complete.

Verification

  • Reproduce the original scroll jank on a release build on a real device, then re-record after the fix; the dropped-frame rate measurably decreases.
  • Insert / remove / reorder operations preserve scroll position and per-item state (focus, expansion, scrubbed video position).
  • Modifier.animateItem() runs the expected fade and placement animation on inserts and removals.
  • Layout Inspector → Recomposition Counts column on item composables plateaus during steady scroll instead of climbing monotonically.
  • @TraceRecomposition on the item composable shows recompositions only on real state changes, not on every scroll-driven invalidation.
  • The compiler report (composables.txt) shows the item composable as restartable skippable with all parameters stable or runtime.

References

Frequently asked questions

What does the Optimizing Lazy Layouts AI skill do?

Use this skill to fix scroll jank, lost item state, and broken animateItem() animations in LazyColumn, LazyRow, LazyVerticalGrid, and LazyHorizontalGrid. Covers stable item keys, contentType for mixed-type feeds, Modifier.animateItem() requirements, hoisting modifier chains and painters out of the items lambda, and validating item composable stability. Use when the developer mentions LazyColumn jank, dropped frames while scrolling, items losing scroll state on insert/remove/reorder, mixed feeds of cards/headers/ads feeling sluggish, animateItem() not animating, RecyclerView view-type analog...

Why use Optimizing Lazy Layouts on TypingMind?

Because you install it once and use it with any model. Optimizing Lazy Layouts is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Optimizing Lazy Layouts in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/skydoves/compose-performance-skills/tree/main/lists/optimizing-lazy-layouts. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Optimizing Lazy Layouts?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Optimizing Lazy Layouts?

As many as you like. As long as a model supports skills, you can use Optimizing Lazy Layouts with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Optimizing Lazy Layouts AI skill free?

Yes. It is published on GitHub by skydoves under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇