Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/ssheleg/super-ux/uxgit clone --depth 1 https://github.com/ssheleg/super-uxWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/ssheleg/super-ux/ux)<a href="https://agentmods.dev/commands/ssheleg/super-ux/ux"><img src="https://agentmods.dev/badge/commands/ssheleg/super-ux/ux.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00050 | $0.03627 |
| Opus 5 | $0.00025 | $0.01813 |
| Sonnet 5 | $0.00010 | $0.00725 |
| Haiku 4.5 | $0.00005 | $0.00363 |
Grade A, and why
ux scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Subtle steeringlowPrompt injection
Instructions that bias recommendations or shape behaviour without the user noticing.
`plans/` dirs) — as `/ux-rule` does. Never tell the user to run a script Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 213 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Single entry point for super-ux. The user is NOT expected to know the skills, layers, or commands inside this plugin — routing is your job. Idempotent: safe to run any number of times, at any project stage.
0. Understand the task (before anything technical)
If $ARGUMENTS already states the task, map it via the routing table and
skip the question. Otherwise ask ONE plain question — "What do you want to
get done with the product's UX?" — with a few examples in everyday words
(new product, new feature, "the UX feels bad", "check that everything
works", "make a fix plan"). Never ask the user to choose between skills or
layers — that vocabulary is internal.
If the task involves designing or improving the interface (routes 1, 3,
4), and the foundation's Design tooling section hasn't recorded a Figma
choice yet, ask one more plain question: "Design the interface visually in
Figma as we go, or text-only? (Figma is the default.)" Record the answer in
foundation.md → Design tooling. If Figma is chosen and the Figma MCP isn't
connected, recommend connecting it and proceed text-only until it is (see
figma-integration reference).
Visual identity, same moment: if screens.md → Design system has no
Style pack and the project has no design system of its own, use the
sheleg-design companion skill to pick one (workbench for product UI /
dashboards / tools, instrument-console, editorial-luxury, or a new pack
on its contract) and record the pack + its token file; a cinematic
scroll-driven landing also takes its motion methodology. Not installed →
offer the one-time install once (/plugin marketplace add ssheleg/sheleg-design-skill + /plugin install sheleg-design@sheleg-design-skill, or npx sheleg-design-skill in the
project) and continue on platform defaults either way — recommend, don't
force. Division of labor: the visual-identity reference.
The second reader, same moment: if the task touches a page a search or an
AI answer engine will read — a landing, pricing, docs, a blog — check
screens.md → Web surfaces. Unanswered, ask once (yes/no), and on yes
every public screen takes the five-field Web surface: block while it is
being designed. This is not an audit item: once a page is live its URL is in
other people's links and its structure is what an answer engine already
quoted. Checking the live page is the seo-aeo-audit companion — not
installed → offer the one-time install once (/plugin marketplace add ssheleg/seo-aeo-audit + /plugin install seo-aeo-audit@seo-aeo-audit, or
npx @ssheleg/seo-aeo-audit in the project) and continue either way.
A brief usually names more than one thing. "A feature, its landing and the funnel" matches three rows below, not one. Map every row that matches, order them by chain position (vision → foundation → flows/screens → scenarios → copy → audit → plan), state the sequence in one line before running the first, and run them in that order. Answering only the row you matched first is how half a brief gets silently dropped.
Routing table (user's words → action from the menu below):
| User says (any language) | Route to |
|---|---|
| "new product", "from scratch" / "стартуем проект", "с нуля" | 1 then 2 then 4 then 6 (vision first, then the chain) |
| "what are we even building", "product vision", "where is this going" / "видение", "что мы вообще строим", "куда идём" | 1 |
| "new feature", "I want to add X" / "новая фича", "хочу добавить X" | 7 (validate idea vs chain — and vs the anti-vision if vision.md exists) then 4/6 for the new parts |
| "the UX is bad", "clunky", improve, redesign / "UX плохой", "неудобно" | 5 (Improve) |
| "check everything works", audit, "run the scenarios" / "проверь что всё работает", "прогони по сценариям" | 8 |
| "what is missing", gaps, coverage / "чего не хватает" | 9 |
| "what do best practices say" / "как лучше по практикам" | 10 |
| "the copy is inconsistent", "tone of voice", "how should this sound" / "текстовка разная везде", "тон оф войс" | 11 |
| "write the button/error/landing/post" / "напиши текст", "перепиши", "заголовок", "лендинг" | 12 |
| "does the copy match the brand" / "проверь тексты по бренду" | 13 |
| "what to fix first", plan / "что чинить в первую очередь", "план" | 14 |
| "the funnel", "pricing page", "checkout", "we need to make money from this" / "воронка", "монетизация", "прайсинг", "оплата" | 4 — a funnel is a flow with screens. The purchase surface in foundation.md routes the practice set (BP-116..123 web, BP-124..129 web2app); then 12 for its copy |
| "design it", "how should it look", "the visuals", "pick a style" / "дизайн", "как это выглядит", "стиль", "визуал" | 4, with the visual identity settled first — the sheleg-design companion picks the style pack and screens.md → Design system records it. Not a taste conversation per screen |
| "will Google/ChatGPT find it", "SEO", "the landing must be findable" / "SEO", "чтобы находилось", "поисковики", "нейронки" | 4 — the Web surface: block is designed with the screen; the seo-aeo-audit companion checks the live page afterwards. Never the other way round |
| "mobile app", "iOS/Android", "which platform" / "мобильное приложение", "айос", "андроид", "приложение" | 2 first — Platform in foundation.md is what routes BP-049..054 and the platform component rules; then 4 |
| "don't know", "just take a look" / "просто посмотри" | run 1–3 of Inspect, recommend from state |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 213 lines · 50 tokens per session scan A 337b2a82ef77
ux is a command published in the GitHub repository ssheleg/super-ux (3 stars, last pushed today), licensed MIT. It adds 50 tokens to every session and 3,627 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (subtle steering). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
test
Run tests and ensure code quality for the Figma Console MCP server.
drift-check
Detect where teams are diverging from the system.
governance-review
Quarterly governance review package.
token-audit
Audit your design token architecture.
component-audit
Audit your component library inventory.
system-benchmark
Benchmark your design system across 12 scored dimensions.