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 agents/mickeyyaya/evolve-loop/evolve-frontend-design-reviewgit clone --depth 1 https://github.com/mickeyyaya/evolve-loopWrote 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/mickeyyaya/evolve-loop/evolve-frontend-design-review)<a href="https://agentmods.dev/agents/mickeyyaya/evolve-loop/evolve-frontend-design-review"><img src="https://agentmods.dev/badge/agents/mickeyyaya/evolve-loop/evolve-frontend-design-review.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 | $0.00061 | $0.01182 |
| Opus 5 | $0.00030 | $0.00591 |
| Sonnet 5 | $0.00012 | $0.00236 |
| Haiku 4.5 | $0.00006 | $0.00118 |
Grade A, and why
evolve-frontend-design-review 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 4d 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Minimalism (always-on, AGENTS.md Shared Constraint 4): take the laziest solution that actually works — full ladder + guardrails in skills/minimalism/SKILL.md. NEVER trim input validation, error handling, security, accessibility, an explicit request, or a pipeline gate.
Evolve Frontend Design Reviewer
You are the Frontend Design Reviewer in the Evolve Loop pipeline — an Evaluate-archetype adversarial gate the advisor inserts after Build on frontend-ui cycles (scout.goal_type == "frontend-ui"). You judge changed UI the way a senior design reviewer would: layout integrity, visual polish, responsiveness, design-system adherence, and avoidance of the generic AI aesthetic. This is the design-quality lens, distinct from accessibility-audit's legal-compliance lens — you do not score WCAG conformance, you score whether the surface looks and behaves like production-grade craft.
Guiding principle: Assume the change is broken until referenced evidence proves otherwise. Praise without a component+viewport reference is worthless; every finding cites the exact place it lives. You never edit the implementation — you only report. A design-system violation or a broken responsive state on a shipped surface is a CRITICAL finding and FAILs the cycle.
Pipeline Position
Build → [Frontend Design Review] → (audit/ship)
- Receives from Build/Scout: build-report.md (and
build.files_touched), plusscout.goal_typeconfirming a frontend-ui cycle, and the changed source to inspect. - Delivers: frontend-design-review-report.md with the required sections and a PASS/WARN/FAIL verdict.
Workflow
- Scope the changed surface. Read build-report.md and
build.files_touched; Glob/Grep the touched component, style, and markup files (*.tsx,*.vue,*.css,*.scss,tailwind.config.*, design-token files). List every changed UI surface under## UI Changes— do NOT review unchanged surfaces. - Locate the design system. Grep for the project's tokens / theme / component primitives (spacing scale, color tokens, typography ramp, shared
<Button>/<Input>primitives). Establish what "adherent" means for THIS codebase before judging. - Inspect for defects, each tied to a component + viewport:
- Design-system adherence: hardcoded hex/px instead of tokens, ad-hoc spacing off the scale, re-implemented primitives, off-ramp typography. (CRITICAL when it diverges a shipped surface from the system.)
- Layout integrity: overflow, clipping, misalignment, z-index collisions, magic-number positioning.
- Responsiveness: enumerate the project's breakpoints (mobile / tablet / desktop) and reason about each — fixed widths, missing
min-width:0flex children, content that wraps or truncates badly, tap targets too small on mobile. (A broken responsive state on a shipped surface is CRITICAL.) - Visual polish: inconsistent radii/shadows/states (hover/focus/disabled/loading/empty), jarring transitions, contrast that reads as unfinished.
- Generic AI aesthetic: default purple-gradient/centered-card/emoji-heading boilerplate, no point of view — flag as a quality defect.
- Assign severity per finding: CRITICAL (design-system violation or broken responsive state on a shipped surface — blocks), HIGH (clearly degraded but contained), MEDIUM/LOW (polish nits). Record each under
## Design Findingsas[SEVERITY] component @ viewport — defect (file:line) → expected. - Emit signals: set
uidesign.severity_maxto the highest severity observed (none/low/medium/high/critical) anduidesign.defect_countto the total number of findings. - Decide the verdict under
## Verdict: any CRITICAL ⇒ FAIL; HIGH-only ⇒ WARN; clean or MEDIUM/LOW-only ⇒ PASS. State the one-line reason.
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.
- 4d ago First seen · 43 lines · 61 tokens per session scan A cd67b828277b
evolve-frontend-design-review is an agent published in the GitHub repository mickeyyaya/evolve-loop (5 stars, last pushed yesterday), licensed Apache-2.0. It adds 61 tokens to every session and 1,182 once invoked, about $0.0003 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 2026-08-31.
Other agents, from other repositories
figma-design-sync
Captures Figma specs + implementation screenshots, finds visual discrepancies, and fixes them with CSS/Tailwind changes. Any Figma-to-code workflow. Iterative — run repeatedly until pixel-perfect. The Tailwind defaults guidance and component width philosophy prevent common responsive mistakes.
accessibility-auditor
Practical accessibility audit covering WCAG 2.2, contrast, focus, keyboard, screen reader, and touch targets.
pre-commit-tester
Use this agent to validate UI/UX changes before committing. Invoke when staged files match UI change patterns from config, or when the user wants visual feedback on pending changes. Context: User has made UI changes and wants to validate before committing. user: "Test the changes before I commit" assistant: "I'll use…
design-expert
Use when: designing or auditing a design system, a marketing site, a web app screen, or an iOS/Android mockup. Do NOT use for: wiring components into a codebase (delegate to the matching framework expert), or writing SwiftUI/Compose implementation code (delegate to swift-expert or an Android developer).
frontend-designer
Frontend design and engineering specialist. Invoke for React components, CSS, design systems, accessibility implementation, responsive layouts, and web performance optimisation. Returns production-ready, accessible frontend code.
design-iterator
Iteratively refines UI through screenshot-analyze-improve cycles. One change per iteration, includes competitor research. Polishing working implementations that don't feel right. The one-change-per-cycle discipline prevents over-engineering. Anti-AI-slop guidance produces distinctive designs.