better
Skill32 starsChanged 24 days agoMITView on GitHub
Improve and implement interfaces across accessibility, layout, writing, typography, colors, and UI polish. Mandatory after building or changing any screen. Use when interface work finishes or a screen or flow must be better, clearer, or more polished; audit only when explicitly requested.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- realdeal88/better-trtheir own version, first seen Jun 8, 2026
- realdeal88/bettertheir own version, first seen Jun 8, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- realdeal88/better★ 0same name
- realdeal88/better-tr★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository42 other files it gives its agents.
- skills/setup-my-hermes-dashboard/template/AGENTS.mdAGENTS.md
- audit-safetySkill
- cleanupSkill
- create-projectSkill
- doneSkill
- dreamSkill
- projectsSkill
- setup-brainSkill
- cloudflareSkill
- code-debugSkill
- codex-imagesSkill
- connect-googleSkill
- connect-instagram-developer-platformSkill
- connect-toolSkill
- connect-vpsSkill
- connect-youtube-developer-platformSkill
- convexSkill
- create-internal-appSkill
- create-landingSkill
- create-softwareSkill
- dev-browserSkill
- grill-meSkill
- initializeSkill
- install-skillsSkill
- lumailSkill
- safe-shipSkill
- setup-cloudflareSkill
- setup-convexSkill
- setup-dev-browserSkill
- setup-githubSkill
- setup-lumailSkill
- dashboardSkill
- setup-my-hermes-dashboardSkill
- setup-nodeSkill
- setup-stripeSkill
- setup-vpsSkill
- show-meSkill
- skill-managerSkill
- stripe-checkoutSkill
- use-styleSkill
- verifySkill
- wait-whatSkill
The file
Files
Improve one resolved screen, flow, feature, or product interface. Default to implementation: inspect, identify the concrete fixes, apply them, and verify the result in the same task. Use the domain guides for design principles and checks; this entrypoint owns workflow, authorization, severity, and reporting.
This is the required last pass for any interface work: run it after building or changing a screen, even when nobody asks. Treat $better, "better this UI", and "make this smarter" as requests to implement. Do not stop after findings, propose work instead of doing it, or ask whether to proceed. Ask only for essential missing information that cannot be resolved from context; continue independent work while waiting. Honor explicit "audit", "review only", "read-only", "plan only", or "no edits" requests as non-mutating boundaries.
For an explicit review of a branch, PR, commit range, or uncommitted changes, review only the interface those changes touch, without editing. A version-control reference in an improvement request identifies the implementation scope; it does not turn the request into a read-only review or require another invocation.
State the scope briefly. Inspect one complete flow at a time; continue through the requested scope rather than silently dropping work. Cover empty, loading, error, and narrow-width states where they exist. Keep a concrete finding list as implementation scope; apply all actionable in-scope fixes, not only HIGH findings. Cap reported findings at 15, not the work performed.
Identify the stack, tokens, viewports, and preview command. Read project interface docs (CONTRIBUTING.md, AGENTS.md, design-system, Storybook). A documented convention does not retire a finding; report a shared-token cause once against that source.
Load the available guides in this order, identify the fixes, then implement:
references/accessibility/guide.mdreferences/layout/guide.mdreferences/writing/guide.mdreferences/typography/guide.mdreferences/colors/guide.mdreferences/ui/guide.md../create-software/references/design-system-rules.mdfor product screens (lists, settings, modals, detail pages). If that sibling skill is not installed, fetchhttps://raw.githubusercontent.com/Melvynx/ai-maxxing-skills/main/skills/create-software/references/design-system-rules.md.
Take principles, references, and verification checks from each guide. This file owns severity, ranking, cap, and format. A missing guide is Not reviewed with its path; do not recreate its rules. Assign an overlapping issue to the owner of the underlying rule and report it once.
Cite path/to/file:line and show the current implementation. Do not report a code finding from appearance alone or a visual finding from source alone when runtime decides the result. Run the project's safe checks; a check you cannot run is Not verified.
HIGH blocks a task, misleads, hides content, risks data loss, or is a confirmed trigger: unnamed control, missing focus, pointer-only path, ignored prefers-reduced-motion, clip at 320px or 200% zoom, failing contrast, color-only meaning, unconfirmed destructive action, unreachable truncated content, undisclosed overflow, error with no recovery, semantic color used against its meaning, motion-only state. MEDIUM harms comprehension, efficiency, adaptability, or consistency. LOW is isolated polish. Rank triggers first, then reach. Prefer delete → platform → existing token → correct value → add. One root cause is one row. Never pad to reach the cap.
Implement with existing components and tokens, preserve unrelated work, and re-verify after changes. When a browser is available, finish with the screenshot loop from the design system rules: capture light and dark, compare side by side with the reference, review bluntly with vision, fix and repeat until no structural issue remains. Distinguish verified behavior from untested states and unrelated check failures. An implementation request does not itself authorize deployment, sending messages, or other external operations.
For implementation, return a concise account of what changed, checks and remaining blockers, with a rendered preview when available. Do not offer to implement work that is already in scope. Use references/review-format.md for explicit review-only requests or when a detailed review is requested alongside implementation. Finish when the requested fixes and feasible verification are complete, or explain the concrete blocker.
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
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.

