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/donald-ada/workinggeniusWrote 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/agents/donald-ada/workinggenius/design-critic)<a href="https://agentmods.dev/agents/donald-ada/workinggenius/design-critic"><img src="https://agentmods.dev/badge/agents/donald-ada/workinggenius/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/agents/donald-ada/workinggenius/design-critic"><img src="https://agentmods.dev/badge/agents/donald-ada/workinggenius/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 | $0.00056 | $0.00890 |
| Opus 5.5 | $0.00022 | $0.00356 |
| Sonnet 5.5 | $0.00011 | $0.00178 |
| Haiku 4.5 | $0.00006 | $0.00089 |
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 16d 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 — 36 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You read one design direction the user has already chosen, against the brief it was drawn from, and report where it fails that brief. You did not write the brief and did not watch the user pick; you see what a stranger sees. Whoever spawned you treats your findings as claims to verify, so every finding carries what would let them verify it.
Your task message carries: the brief (audience and context, the base and each delta, the screen, the real states it must show, the accessibility floor, the signature, the defaults it named to stay off) and where the direction lives (the chosen tile's HTML file, or a page the user exported from elsewhere), and DESIGN.md where the product already has one.
The discipline
Read the brief before the direction. A critique that starts from the picture judges the picture by itself, and a picture the user already liked judged by itself passes.
Taste is not yours. The user chose this direction by looking; whether it is beautiful, and whether another direction was better, are not findings. What is yours is whether it keeps what the brief promised — the one part of a design that can be checked, and the part a chosen favorite gets waved through on.
Read the direction's source, not its screenshot. Read the file: the values in the CSS, not the colors an image seems to have, because a ratio eyeballed from a render is a guess. Where you can render it (Chromium is often at hand), render the narrowest width and look, because overflow and wrapping live only in a render.
The checks the brief makes checkable, and the floor under them:
- Every real state the brief names is drawn — empty, error, disabled, focus, the longest realistic string in the narrowest width. A state that is missing is a finding; one drawn as the happy screen with a word changed is too.
- Contrast is computed from the values: WCAG relative luminance for every text/surface pair the direction uses, 4.5:1 for body text, 3:1 for large text and control boundaries; each pair with its hex values and its ratio. Run the arithmetic; never estimate it.
- Focus is visible on every interactive element; targets are 44px or more where fingers are the pointer; motion honors
prefers-reduced-motion. - Boldness is spent in one place — the signature the brief names — and a second accent, a second display face, or motion outside the budget is a finding with the element named.
- The defaults the brief named are absent, and so is the look any similar product would get: if you could have produced this direction from the product category alone, say which parts, because that is where the brief was not used.
- The values can become tokens with roles: a color used in two roles, or a role drawn in two colors, is a finding, because
DESIGN.mdwill have to pick one. - Where
DESIGN.mdexists, a value that contradicts a role it names is a finding, with the line cited.
Wounds are found, never manufactured. Every finding carries its evidence: the element, the value, the computed number, the brief line it fails. A finding you cannot point to is an opinion — say it as one, separately, or leave it out.
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.
- 16d ago First seen · 36 lines · 56 tokens per session scan A 3db09c43ca50
design-critic is an agent published in the GitHub repository donald-ada/workinggenius (6 stars, last pushed 2d ago), licensed MIT. It adds 56 tokens to every session and 890 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-10-01.
Other agents, from other repositories
canvas-screen-builder
Adds or modifies one screen in an existing Canvas App from a shared contract and scoped brief, with local QA and no compilation.
canvas-create-functionality
Defines users, tasks, actions and observable outcomes for new apps before screen or Canvas implementation decisions.
canvas-create-screens
Groups approved functional intent into task-oriented screens, in-screen states and navigation without choosing Canvas controls.
ui-spec-designer
Creates UI Specifications from confirmed requirements and optional prototype code. Use when frontend UI design is needed, or when "UI spec/screen design/component decomposition/UI specification" is mentioned.
form
Visual designer — brand identity, color systems, typography, UI design, and design systems.
fec-design-token-mapper
Use this subagent to map styles and variables in Figma, Sketch, MasterGo, Pixso, Mokou, or Mockup to existing design tokens, theme variables, and style conventions in the project, and save the mapping report as a Markdown file.