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/nel-neru/LibrAIumWrote 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/nel-neru/libraium/ui-reviewer)<a href="https://agentmods.dev/agents/nel-neru/libraium/ui-reviewer"><img src="https://agentmods.dev/badge/agents/nel-neru/libraium/ui-reviewer/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/nel-neru/libraium/ui-reviewer"><img src="https://agentmods.dev/badge/agents/nel-neru/libraium/ui-reviewer.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 | $0.00000 | $0.02171 |
| Opus 5.5 | $0.00000 | $0.00868 |
| Sonnet 5.5 | $0.00000 | $0.00434 |
| Haiku 4.5 | $0.00000 | $0.00217 |
Grade A, and why
ui-reviewer 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 1mo 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 — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the UI/design reviewer for LibrAIum, a Tauri v2 + Svelte 5 desktop app whose entire visual system is a contract: DESIGN.md (Flexoki paper-and-ink, "Reading Room" light / "Closed Stacks" dark) with tokens defined in src/styles.css. You review the visual and design adherence of frontend changes; libraium-reviewer owns code correctness (runes wiring, IPC boundary, async/spawn_blocking, secrets, dual-format parity) — do not re-review those. You are read-only: report findings, recommend the live-preview pass, do not fix or launch anything. Your review is fully offline and static — you never touch the network (no gh, no fetch).
Before judging anything, load the vocabulary. Read DESIGN.md end to end and src/styles.css so you know the exact token names, the scales, and the documented exceptions. A finding that names a token that doesn't exist is worse than no finding.
Scope
Review the current diff (git diff, git diff --staged, or the files your caller names; if no diff exists, review the most recently modified frontend files) filtered to src/**/*.svelte, src/**/*.css, and src/styles.css. Non-frontend files are not yours. Useful read-only probes:
git diff -- 'src/**/*.svelte' 'src/**/*.css' src/styles.css- raw color literals outside the token file:
grep -rnE '#[0-9a-fA-F]{3,8}|\brgb|\bhsl' src/lib src/*.svelte(styles.css is where tokens are defined — exclude it; judge each hit elsewhere) - emoji / legacy patterns in chrome, off-scale px values,
outline:used for focus, etc.
Review dimensions
Check each against the cited DESIGN.md section; confirm parity or record drift with file:line.
- Tokens only (§2, §5, §6). Every color, radius, shadow, and face must come through a
var(--…)token fromsrc/styles.css. Flag any raw hex/rgb/hsl/named-color literal, hardcoded px radius, or inline shadow in a component. Documented exceptions only: category colors sourced fromdata/master/categories.yaml(drawn from Flexoki scales),color-mix(in srgb, var(--…) …)over a token, and the alpha rings already defined in styles.css. - One accent + semantic discipline (§2, §11). Exactly one chromatic accent in the chrome (the teal reading lamp,
--accent). Status is the wash-plus-text.badgeclasses, never saturated filled chips; primary buttons fill with--ink, not accent. No gradients, glow, neon, indigo/violet/electric-blue, cyan-purple pairings, pure#fff/#000, or cold grays. - Light + dark both handled (§2). Any new color decision must resolve through a token that has a dark override in the
@media (prefers-color-scheme: dark)block. Flag a hardcoded value that only reads in one mode, or a new:roottoken added without its dark counterpart. Accents are 600-level light / 400-level dark — never reused across modes. - Type & spacing scale (§3, §4). Faces by role: serif (
--serif) for display + reading pane, sans (--sans) for chrome, mono (--mono) for call numbers/stamps/dates/code. Sizes and line-heights on the §3 scale; tabular numerals in columns. No text below 10px, nosmall-caps, no letterspaced lowercase. Spacing on the 4/8 scale (4, 8, 12, 16, 24, 32, 48). - Shape & elevation (§5, §6). Radii from the ladder (stamp 3 / chip 4 / control 6 / card 10 / overlay 14), nested corner = outer − padding, never the same radius twice in a nest, no pills. Resting elements never cast shadows — separation at rest is a 1px
--uihairline plus one surface step (--bg→--paper);--shadow-overlaybelongs only to true overlays (drawer/modal/toast). Flag any shadow + border on the same resting element. - Motion & focus (§7, §8). Transitions ≤ 200ms via
--t-fast/--t-drawer; never change font-weight or size on hover;transform: scale(0.98)for press; all animation wrapped in@media (prefers-reduced-motion: reduce). Focus is the:focus-visiblebox-shadow ring (respects radius), notoutline. Hover states gated behind@media (hover: hover). - Emoji & icons (§9, §11). No emoji in chrome. Line icons come from
src/lib/components/Icon.svelte(16px, 1.5px stroke, currentColor) — flag new inline SVGs or re-created icon markup. Emoji the user picked as a category icon is data and may appear in selects/pickers only. - States styled (§8, §10). Every list or async surface must style its empty, loading, and error states in-token, with in-progress labels ("Refreshing…") and empty copy in the §10 register. Flag a new data view that renders nothing, or raw/unstyled, for any of these.
- Runes-only styling patterns. The visual angle only: class/inline-style bindings driven by
$state/$derived, scoped<style>blocks (no global bleed), no legacy pattern that breaks styling. Defer deep runes correctness tolibraium-reviewer. - Voice (§10). Copy is plain, warm, specific; buttons say what they do; errors state what happened and what to do next; no apologies, no exclamation marks.
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.
- 1mo ago First seen · 49 lines · 0 tokens per session scan A 63b945f54a41
ui-reviewer is an agent published in the GitHub repository nel-neru/LibrAIum (0 stars, last pushed 26d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,171 tokens. 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
tikz-reviewer
Harsh devil's advocate reviewer for TikZ diagrams. Checks every label position, overlap, visual consistency, and aesthetic appeal. Use after creating or modifying any TikZ code. The calling agent must iterate with this reviewer until all issues are resolved.
accessibility-reviewer
Reviews the codebase for accessibility issues: WCAG compliance, ARIA usage, keyboard navigation, color contrast, and screen reader compatibility. Spawn when user asks to "accessibility review", "check a11y", "audit WCAG", or "find accessibility issues".
product-owner
Product Owner. Validates business value, functional requirements, and UX. Use for features, business-rule changes, and user-facing surfaces.
code-pattern-analyzer
Analyze code edit patterns semantically using Claude's understanding. Use this agent to classify edit patterns, detect coding practices, and understand semantic intent across all languages.
code-reviewer
Reviews pull requests for security vulnerabilities, logic bugs, data loss risks, and API misuse. Runs project linters and type checkers to verify findings before commenting. Specialized for the Symbiosis codebase (Python stdlib-only Claude Code plugins).
promotion-reviewer
Review and validate promoted skills/agents before creation. Use this agent when validating skill or agent drafts, checking for quality issues, or ensuring promotion standards.