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 skills/aiocean/claude-plugins/aio-visual-diffnpx skills add aiocean/claude-plugins --skill aio-visual-diffgit clone --depth 1 https://github.com/aiocean/claude-pluginsWrote 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/skills/aiocean/claude-plugins/aio-visual-diff)<a href="https://agentmods.dev/skills/aiocean/claude-plugins/aio-visual-diff"><img src="https://agentmods.dev/badge/skills/aiocean/claude-plugins/aio-visual-diff.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 | $0.00116 | $0.02924 |
| Opus 5 | $0.00058 | $0.01462 |
| Sonnet 5 | $0.00023 | $0.00585 |
| Haiku 4.5 | $0.00012 | $0.00292 |
Grade A, and why
aio-visual-diff 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 yesterday.
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 — 232 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Visual Diff — Measurement Driven, Not Eyeballed
"Heading font size is 28px, spec says 24px, change Heading size from 7 to 6." Specific, numerical, unambiguous. That is feedback the agent can act on.
This skill drives a measurement loop: navigate the dev server, extract computed CSS and bounding-box numbers via chrome-devtools MCP, diff against a Figma node (Figma fidelity mode) or a frozen baseline JSON (baseline regression mode), then convert every delta into a concrete "change X to Y in file Z" correction. The loop repeats until all deltas fall within threshold — no screenshot eyeballing at any step.
LLM (kể cả multimodal) yếu spatial pixel reasoning — đưa nó 2 screenshot rồi hỏi "giống không?" sẽ trả lời chủ quan và thường sai. Skill này thay bằng vòng lặp đo số → diff số → fix theo delta.
Required Setup
chrome-devtoolsMCP (bắt buộc) — extract measurements từ DOM thật. Tools dùng:navigate_page,evaluate_script.figmaMCP (optional) — Figma fidelity mode. Tool dùng:get_design_context. Không có → tự động fall sang baseline regression mode.
Cả hai đều phải có dev server đang chạy ở URL truyền vào.
Two Modes
| Mode | Khi dùng | Reference source |
|---|---|---|
| Figma fidelity | Có Figma file làm spec | figma__get_design_context → node layout numeric |
| Baseline regression | Không có Figma, chỉ chống drift | .aio-visual-diff/<selector-hash>.json lần đo đầu tiên |
Workflow
Step 1 — Identify target
Hỏi user chỉ những gì còn thiếu:
- Dev URL (vd
http://localhost:3000) - Selector của component cần verify (xem Selector Strategy bên dưới)
- Figma node URL (optional) — không có → chạy baseline mode
Step 2 — Render & extract numbers
chrome-devtools__navigate_page → <dev-url>
chrome-devtools__evaluate_script → measurement function (paste-as-string)
Measurement function template — replace SELECTOR_HERE, paste nguyên cả block làm function arg cho evaluate_script:
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.
- yesterday First seen · 232 lines · 116 tokens per session scan A 4f1407822ccb
aio-visual-diff is a skill published in the GitHub repository aiocean/claude-plugins (4 stars, last pushed 2d ago), licensed MIT. It adds 116 tokens to every session and 2,924 once invoked, about $0.0006 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 2026-09-03.
Other skills, from other repositories
landing-page-generator
Generate complete, deploy-ready landing pages from any repository. Use when creating a homepage for an open-source project, building a project website, converting a README into a marketing page, or standardizing landing pages across multiple repos.
product-ui
Use this skill when UI design and UI implementation should happen together. It turns product context into usable frontend screens, components, layout decisions, design-system choices, accessibility rules, and visual polish.
ui-web
Web UI - glassmorphism, Tailwind, dark mode, accessibility.
panel-app-react-vite-creator
Create or update an engineering-style NextClaw Panel component with pnpm, Vite, React, TypeScript, and Tailwind CSS, then build it into a static .panel directory inside a schema v2 package or an explicitly loose workspace Panel. Use for modern, reusable, complex, Agent-powered, typed Panel interfaces.
eval-agents
Audit Claude Code agents defined in .claude/agents/ for description specificity, model tier appropriateness, tools scoping, and system prompt quality. Detects dispatch ambiguity between agents, flags over-permissive tool grants, and checks for human-in-the-loop patterns that break programmatic orchestration. Use when…
land-and-deploy
Merge PR, wait for CI, verify deploy, run canary. The complete landing pipeline.