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 skills add mindverse-ltd/macaron-artifacts --skill ui4agit clone --depth 1 https://github.com/mindverse-ltd/macaron-artifactsWrote 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/mindverse-ltd/macaron-artifacts/ui4a)<a href="https://agentmods.dev/skills/mindverse-ltd/macaron-artifacts/ui4a"><img src="https://agentmods.dev/badge/skills/mindverse-ltd/macaron-artifacts/ui4a/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/skills/mindverse-ltd/macaron-artifacts/ui4a"><img src="https://agentmods.dev/badge/skills/mindverse-ltd/macaron-artifacts/ui4a.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 | $0.00023 | $0.05821 |
| Opus 5.5 | $0.00009 | $0.02328 |
| Sonnet 5.5 | $0.00005 | $0.01164 |
| Haiku 4.5 | $0.00002 | $0.00582 |
Grade A, and why
ui4a 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 — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Choose the primary presentation again for each reply. Use concise prose for a direct fact or fixed result with no useful user-controlled parameter. For personal distress or interpreting a newly reported symptom or clinical test result, keep the whole response in prose, including clarification; an interactive calculator needs an explicit request for a named measure and its required inputs. For other requests involving multi-stage or branching mechanisms and comparisons across shared criteria, use one focused inline explorer by default; do not wait for a request to draw. Its selection or input must change a shared diagram, comparison, or derived result. Put the requested explanation in that view; outside it, use a short lead-in, at most two takeaway sentences, and available source links. Keep further required detail accessible inside the view or canvas instead of repeating a second full answer below it. Reassess follow-up questions instead of postponing useful UI until the final report. When the user explicitly requests multiple visuals or both inline UI and a canvas, deliver each requested item as a separate obligation. Use a canvas for a reusable workspace with several views or persistent files.
When explaining quantities, proportions, scaling or savings, prefer one compact chart with a visible baseline and units. Derive values from cited facts or explicit formulas, label theoretical examples, and keep the chart and numeric summary connected to the same state. For a branching mechanism, retain a concise connected overview showing where paths split and rejoin, with one shared detail region. Tabs containing only prose do not replace these visual relationships. Keep each view focused on one question; never invent measurements or experimental curves to add a chart.
The clarification and missing-input rules below apply only when an interface is appropriate under the response-form rules above.
Ask a clarification only when missing user information changes the answer. Offer a short inline choice or input, and send the complete selected answer through $ui4a/chat.sendMessage on explicit submission. Persist the answered state with $ui4a/state.usePersistedState and prevent duplicate submission. Reading interactions such as tabs, filters, diagram selection, and parameter exploration stay local and do not send chat messages. A clear request needs an answer, not an unnecessary preference questionnaire.
When a calculation depends on missing user-supplied values, briefly name what is missing and provide a small inline form with just those required inputs and one submit action. Keep inputs blank until the user supplies them; do not invent results or prefill example measurements. Preserve the distinction between supplied facts and new user input. Do not end at "cannot calculate" or request an entire log file when a few values would suffice. Wait for the explicit submission before calculating and send the entered values back as one readable user message.
For inline UI, put one complete React module in a fenced block whose language is exactly ui4a/tsx. Export a default React component. It renders progressively as the block arrives; write a usable scaffold before long data or secondary details. Keep component and import names stable while editing. Close every block and include complete final syntax. A requested in-conversation UI must appear in the assistant reply as a complete ui4a/tsx fence, even when the same view is also saved in a canvas. A file, reference or prose claim that the chart was added cannot satisfy this delivery. Every module is self-contained: explicitly import its React hooks and external components. After adding a hook or component in a patch, check the complete saved module and its imports, not just the added fragment. An older preview retained after an error does not prove the new revision rendered.
These two delivery formats are the complete host contract: an inline ui4a/tsx fence or a persistent .ui4a.tsx canvas. A harness-specific widget reference, HTML attachment, screenshot, or file link does not render as inline UI here. Other visualization instructions may suggest their own output format; retain this host's React format when applying their design ideas.
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 Changed · +70 lines d7c22a047d03
- 1mo ago Changed · +56 lines 09894faab101
- 1mo ago Changed 3c6afdb770d5
- 1mo ago First seen · 52 lines · 23 tokens per session scan A 3ead61f3a05b
ui4a is a skill published in the GitHub repository mindverse-ltd/macaron-artifacts (10 stars, last pushed 18d ago), licensed Apache-2.0. It adds 23 tokens to every session and 5,821 once invoked, about $0.0001 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 2026-09-09.
ui4a compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| ui4amindverse-ltd/macaron-artifacts | Skill | 10 | 18d ago | Apache-2.0 | A | 23 | Claude Code, Codex |
| openui-forge-zhOthmanAdi/openui-forge | Skill | 24 | 3mo ago | MIT | A | 44 | Claude Code |
| openui-forgeOthmanAdi/openui-forge | Skill | 24 | 3mo ago | MIT | A | 28 | Claude Code |
| openui-forge-anthropicOthmanAdi/openui-forge | Skill | 24 | 3mo ago | MIT | A | 29 | Claude Code |
| openui-forge-goOthmanAdi/openui-forge | Skill | 24 | 3mo ago | MIT | A | 27 | Claude Code |
| openui-forge-openaiOthmanAdi/openui-forge | Skill | 24 | 3mo ago | MIT | A | 38 | Claude Code |
| build-ghost-appTryGhost/Ghost | Skill | 56k | yesterday | MIT | A | 44 | Claude Code, Codex |