agentleFS
Sign inStart free
obakeng-developsobakeng-develops/taste/skills/taste-sweep/SKILL.md

taste-sweep

Skill0 starsChanged 5 months agoMITView on GitHub

Sweep a whole codebase for design and prose taste, not one screen. Use when the user says "taste sweep", "taste-sweep", "audit the codebase for taste", "audit the repo for design", "where's the jank", or "find taste debt". Produces a ranked report of what to fix across the front-end and user-facing text. Read-only — applies no fixes. For a single screen or passage, use /taste-audit instead.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the taste library at ~/.claude/taste/references.md (if it exists).

    line 11
  2. Distinctive:

    Don't read every file. Find the surface, then sample what carries the most weight.

    line 16
  3. Distinctive:

    Use Explore/Grep to locate.

    line 22
  4. Distinctive:

    No fixes applied. Skip empty buckets.

    line 49

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Taste — sweep

Audit a whole codebase for taste. Find the patterns, not one-off nits. Report only.

Calibrate first

Read the taste library at ~/.claude/taste/references.md (if it exists). Judge against the user's taste, not a generic ideal. A finding that fights a saved reference loses.

Map the surface

Don't read every file. Find the surface, then sample what carries the most weight.

  • Design system: tokens, theme, global CSS, the shared component library. Drift here multiplies everywhere — start here.
  • High-traffic screens: entry points, the main flow, anything money- or auth-shaped.
  • User-facing text: button labels, empty/error/success copy, headings, toasts, validation messages — grep the string surface. Use Explore/Grep to locate; for a large repo, fan out with subagents (or a workflow if the user opts in). Sample representative files; say what you sampled and skipped.

What to look for — patterns, ranked by blast radius

  1. Systemic jank: more than one typeface family; ad-hoc spacing/colors instead of tokens; inconsistent radius, shadow, or alignment; body text outside 15–25px / line-height 1.3–1.45 / 45–90 chars.
  2. Affordance rot: wrong control for the job repeated across screens; conventions broken by accident; inconsistent interactive states (hover/focus/disabled).
  3. Story gaps: missing empty/error/success states; hierarchy that buries the user's real question; flows with no arc.
  4. Prose clutter (Zinsser): copy thick with dead words and qualifiers; passive voice and weak verbs in UI text; jargon; stiff, voiceless microcopy.

Report

Open with a one-line verdict on the codebase's taste health. Then findings ranked by blast radius — widest first — one line each: pattern, where it lives (file:line or "across N components"), the fix. Group as:

  • Systemic jank / clutter — fix first; one change, many screens.
  • Affordance / clarity rot
  • Story / voice gaps
  • Polish Close with the single highest-leverage change and a rough count of what you didn't sample.

Rules

  • Hunt repeated patterns; a one-screen nit belongs in /taste-audit, not here.
  • Name the mechanism and the blast radius, not "feels off".
  • No fixes applied. Skip empty buckets. Flag what you sampled vs skipped — silent truncation reads as "covered everything" when it didn't.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free