Labsco
Labsco/Repos/sveltejs/svelte
REPO PACKAGE

sveltejs svelte

COMMUNITY
sveltejs · publisher87,507 repository starsMIT · Freegithub.com/sveltejs/svelte
1skill
0ready to use

LABSCO SUMMARY

performance-investigation wraps the Svelte monorepo's own benchmark tooling. It runs pnpm bench:compare main <branch> to check out both branches, run the benchmark suite against each, and write a diff report plus per-benchmark CPU profiles; the suggested flow is to open the report, find the largest regressions, then compare the hotspot summaries for each branch against Svelte's own runtime internals — runtime.js, reactivity/batch.js, reactivity/deriveds.js, reactivity/sources.js. A helper script, profile-diff.mjs, prints the top function-level sample-share deltas between two branches for one benchmark, and the skill points you at pnpm test runtime-runes to re-check correctness after a perf change.

This is for someone already working inside the sveltejs/svelte repository who suspects a change slowed down reactivity or compilation, not for someone writing components or debugging performance in their own Svelte app.

READ THE FULL ANALYSIS

It runs destructively against your working tree. bench:compare checks out both branches as part of the run, and its own documentation warns to commit or stash first; each run also overwrites the previous .results and .profiles output directories rather than keeping history.

This is the entire agent-skill surface Svelte ships. There is no skill here for writing components, using runes, or anything else in the framework itself — just this one internal benchmarking aid, bundled in the same repository as the compiler.

Checked 19 September 2026 against the skill's SKILL.md body; the repository README does not mention the skill at all.

87,507Stars on the GitHub repository, at last check.