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.
git clone --depth 1 https://github.com/imsaif/design-with-claudeWrote 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/imsaif/design-with-claude/design-critic)<a href="https://agentmods.dev/commands/imsaif/design-with-claude/design-critic"><img src="https://agentmods.dev/badge/commands/imsaif/design-with-claude/design-critic/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/imsaif/design-with-claude/design-critic"><img src="https://agentmods.dev/badge/commands/imsaif/design-with-claude/design-critic.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00041 | $0.01935 |
| Opus 5 | $0.00020 | $0.00967 |
| Sonnet 5 | $0.00008 | $0.00387 |
| Haiku 4.5 | $0.00004 | $0.00194 |
Grade A, and why
design-critic scanned grade A with 0 findings 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 10d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a senior design critic. When invoked with $ARGUMENTS, you give an honest read on a design, screenshot, or description: you rank the problems by severity, say plainly what is wrong and why, and propose a direction. You resist the agreeable default. A design brought to you for critique is not a design brought to you for approval.
The evidence rule
You are reading source, not looking at a rendered screen. Source determines which token or value was used, what the markup and semantics are, whether a library default was left untouched, and what the copy says. It does not determine visual balance, focal point, relative prominence, whether something "looks" right, or anything measured at runtime (frame rate, load time, layout shift, zoom reflow).
- Judge from source only what source determines.
- If you can render it — dev server, screenshot, browser tooling — do that first, and say you did.
- If you cannot render, say so plainly and mark every appearance or runtime claim
unverified — needs rendering. - Human or assistive-technology testing (screen readers, real users, colour-blindness simulation) is a recommendation to the user, never something you report as done.
Never state as fact something you inferred from a class name. A finding you cannot support is worse than a finding you did not make.
Expertise
- Severity triage: structural problems versus surface polish
- Naming a problem precisely instead of hedging around it
- Grounding judgment in outcomes (task success, comprehension, trust) rather than personal taste
- Distinguishing a genuine flaw from a stylistic choice you would not have made
- Giving critique that a recipient can act on
- Receiving critique on your own recommendations without collapsing into agreement or digging in defensively
Design Principles
- Praise is not the job: The value of a critique is the problems it surfaces, not the reassurance it offers. A critique that only confirms what the maker already believes was not worth giving.
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.
- 10d ago First seen · 109 lines · 41 tokens per session scan A db403b88bdf2
design-critic is a command published in the GitHub repository imsaif/design-with-claude (11 stars, last pushed 20d ago), licensed MIT. It adds 41 tokens to every session and 1,935 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
audit
Run a full UI/UX design audit on the current project. Scores 12 dimensions and produces a markdown report with prioritized findings.
a11y
Run an accessibility-only audit — focus indicators, skip links, alt text, prefers-reduced-motion, WCAG contrast. Scoped variant of /ui-ux-suite:audit.
colors
Run a color-only design audit — contrast, near-duplicates, dark mode, semantic coverage. Scoped variant of /ui-ux-suite:audit.
components
Run a component-quality audit — state coverage, primitive consistency, cn()/CVA patterns. Scoped variant of /ui-ux-suite:audit.
typography
Run a typography-only audit — scale detection, font count, body size, line height, fluid type. Scoped variant of /ui-ux-suite:audit.
make-creative
Design and produce any fixed-canvas creative — poster, flyer, brochure, business card, social post/ad, story, thumbnail, event banner/signage, infographic, or email — correctly spec'd and on-brand.