Index of every page on this site: llms.txt
All skills
vercel-labs avatar

/ui-sdk-unslop

@fb6945a
by Vercel Labsvercel-labs/ui6.3k stars
218
Run it yourself

Removes generated-looking code from a ui-sdk diff or files, like comments that restate the code, filler JSDoc, defensive checks for impossible cases, one-caller helpers and vague names, without changing behavior. Use as the last step of any change, or when asked to clean up, trim comments or remove slop.

  • 1 file
  • 4.3 KB
  • Updated 2 weeks ago
  • GitHub
Use this Skill: https://skilld.dev/gh/vercel-labs/ui/ui-sdk-unslop

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈80 tokens for metadata: the name and description. ≈1k when used: this file.

Description uses 4.0% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈321 characters, or 4.0%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Unslop

Scope is the diff (git diff against the base) unless you're given files. Behavior never changes: the same tests pass before and after, and nothing public is renamed.

What goes

Each pattern, with what replaces it.

A comment that says what the next line does. Delete it. If the line isn't clear without it, rename something.

// Close the popup when Escape is pressed.
if (event.key === "Escape") close()

A comment on every declaration. Most constants, fields and functions need none: the name and the type say what they are. Delete each one that doesn't stop a reader from getting the code wrong.

A comment that explains the design. What a value holds, where it's kept, which attribute it lands on, who reads it. The code says that. Keep the one fact it can't show, in a line, or delete the comment.

// Before
// The site's color scheme, shown on <html> as `data-color-scheme`, which the
// stylesheet's colors follow. It's kept under "theme", the key the site used
// before, so readers keep what they chose.
const colorScheme = createColorScheme({
  storage: createLocalStorage("theme"),
  global: true,
})

// After
// "theme" is the key the site used before, so readers keep their choice.
const colorScheme = createColorScheme({
  storage: createLocalStorage("theme"),
  global: true,
})

A comment longer than three lines. Cut it to the reason. An explanation that needs more goes in the docs page, a test title or contributing/. pnpm lint flags it (comment-length).

A comment that narrates history or the task. "Now we", "Updated to", "Fixed the bug where", "As requested". Delete it. Git holds the history.

A section banner. // ---- Helpers ----, // State. Delete it. Split the file if it needs signposts.

JSDoc that repeats the name. /** The value. */ value. Say what the reader can't see: the default, the unit, when it's called, what happens on change. pnpm lint catches field docs that only repeat the name; write the rest the same way.

A hedge or a sales word in a comment or a doc. "Robust", "seamless", "comprehensive", "simply", "just", "easily", "ensure", "leverage", "utilize", "note that", "it's important to". Cut the word; if nothing is left, cut the comment.

A defensive check the types already rule out. if (!items) return on items: Item[], ?? [] on a value that's never undefined, try around code that can't throw. Delete it. Narrow unknown at a boundary instead.

A cast or ! with no reason. Make the type right, or say why in a comment.

A one-caller helper, or a wrapper that only renames. Inline it, unless its name explains something the call can't.

A name that says nothing. data, result, item2, handleClick2, temp, flag, isValid2. Name what it is: highlightedItem, isOpen, onTriggerPress.

An option nobody passes, a parameter with one value, a config object for one setting. Delete it.

Logs, probes and commented-out code. Delete them.

In examples, any comment at all. Only tool directives stay (svelte-ignore, eslint-disable, @__PURE__).

What stays

  • A comment that says why, in three lines at most: a browser bug and the browser, an APG rule, a trade-off, a link to the spec or test that pins the behavior.
  • The word "primitive", and parentheses that hold code names, like "the trigger (MenuTrigger)".
  • JSDoc on every public field and export. Make it say more; don't delete it.

Steps

  1. List what you'll change, by file, before touching anything.
  2. Make the changes.
  3. Run pnpm lint, pnpm format, pnpm typecheck and the unit tests for the packages you touched. Check exit codes.
  4. Delete each comment left in the diff and reread the code. If it still reads right, leave it deleted.
  5. Read the diff once more. Revert any change that doesn't make the code easier to read.
  6. If one pattern showed up in three or more files, propose a lint rule for it in scripts/check-code.mjs or scripts/check-comments.mjs.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at fb6945a. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 14 hours ago.

Activeupdated 2 weeks ago
metadata
{
  "internal": true
}

README badge

README badge for vercel-labs/ui/ui-sdk-unslop