Borrowing it
Nothing to install: this file belongs to miikkij/aimeat-protocol. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/miikkij/aimeat-protocol/main/.claude/skills/aimeat-frontend-verify/SKILL.mdgit clone --depth 1 https://github.com/miikkij/aimeat-protocolWrote 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/miikkij/aimeat-protocol/aimeat-frontend-verify)<a href="https://agentmods.dev/skills/miikkij/aimeat-protocol/aimeat-frontend-verify"><img src="https://agentmods.dev/badge/skills/miikkij/aimeat-protocol/aimeat-frontend-verify/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/miikkij/aimeat-protocol/aimeat-frontend-verify"><img src="https://agentmods.dev/badge/skills/miikkij/aimeat-protocol/aimeat-frontend-verify.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.03871 |
| Opus 5.5 | $0.00000 | $0.01548 |
| Sonnet 5.5 | $0.00000 | $0.00774 |
| Haiku 4.5 | $0.00000 | $0.00387 |
Grade A, and why
aimeat-frontend-verify 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 today.
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 — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verifying a finished frontend change
Do not write or run the .spec.ts Playwright suite (pnpm test:playwright:*). It is unreliable. Drive a real browser through the Playwright MCP server (.mcp.json) instead.
Trigger: a completed change to public/views/, public/components/, public/js/, public/css/, locales/, or a *.html page, or a published single-file AIMEAT app (aimeat_app_publish). Mid-development is not the trigger; done is.
Against your own sandbox node (pnpm sandbox, port 40600 upward, owners and passwords in aimeat/.sandbox.json; restart it to pick up a backend change), not the shared dev server on 40050, which belongs to the developer and may be restarted by another session: navigate to the page, reach the authenticated state, perform the real interactions, and confirm the expected result actually happens. Elements appear, data persists, edits and deletes take effect.
A published app is checked signed in, in the sandbox, before anyone sees it in production. Publish it to the sandbox (POST /v1/apps with an owner token from .sandbox.json) and open <baseUrl>/v1/apps/<owner>/<file>?mode=inline: it redirects to the app's own origin, <name>.apps.localhost:<port>, where the pill's Sign In opens the same grant window as on aimeat.io; sign in there with a sandbox owner. Until 2026-09-29 the sandbox ran apps in an opaque frame whose sign-in the browser refused, and Postinjalostamo's corrections were found by Jouni in production instead.
An HTML page of a workspace is checked in its frame, in the sandbox. The sandbox runs the page host at pages.localhost:<port>, as aimeat.io runs it at pages.aimeat.io: open the page in the workspace view (<baseUrl>/v1/profile?tab=organisms), or share and publish it and open its viewer link. The page loads only inside the frame; opened in a tab of its own it goes to the viewer, so a direct load of a pages.localhost address proves nothing.
An app that reads mail or asks a model is checked on the sandbox's stand-ins. The first owner is connected to a test mailbox (19 sample messages, a PDF invoice with a text layer and a scanned one, and one request for a quote that a definition with its own class quote sorts; the connection id is mailbox in .sandbox.json), and the node's decision model and completion model are local stand-ins that answer each sample with the class and fields a person would give it. A batch, a classification or an extraction therefore runs in the sandbox with no paid key, and its result is known before you look. Speech to text has a stand-in too: every recording is answered with one short Finnish meeting note, so a recorder's path from audio to text can be run. Sending mail has none: the sandbox sends nothing, and a send answers that mail is off. A spent AI allowance is seen by starting the sandbox with AIMEAT_SANDBOX_FREE_MODEL=openrouter/free (and AIMEAT_GOOSE_BIN for the chat) and spending the stand-in's key with POST http://127.0.0.1:<port+100>/sandbox/ai-key {"limit":0.01,"usage":0.02}; {} takes the limit away again.
A page about agents and the machines that run them is checked with pnpm sandbox:machines. It leaves two stand-in machines on the sandbox while it runs (one connected, one away), agents on them that work by the clock, when asked and when written to, one agent that waits for the absent machine and one proposal that waits for the owner, and it puts the first owner on the finished home. The completion stand-in also answers the home's "new agent" sentence with a draft. Sign in with the password, as pnpm sandbox does for every owner: its tokens come from the password sign-in and last a day, so a browser sign-in in between changes nothing for them. A page as it reads on a node that takes no computers (the connector tunnel off, as on innokas.aimeat.io on 2026-10-14) is checked on a sandbox started with AIMEAT_SANDBOX_TUNNEL=false pnpm sandbox.
An app built from an approved drawing is compared with it before Jouni sees it: pnpm ui:compare. Publish the approved genre (or the page he approved) to the sandbox beside the app, then, from aimeat/:
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.
- today Changed bf4f181510f8
- 13d ago Changed · +2 lines 0103558bfaa2
- 15d ago Changed 623aa162dc05
- 16d ago Changed c5562889927d
- 17d ago Changed a47bd6500da5
- 19d ago Changed · +2 lines 91d5fcdded7d
- 20d ago Changed 0dded7d945f2
- 22d ago Changed 4bba180e4367
- 1mo ago Changed · +13 lines 8a86c380095f
- 1mo ago First seen · 117 lines · 0 tokens per session scan A 11038fe4b626
aimeat-frontend-verify is a skill published in the GitHub repository miikkij/aimeat-protocol (10 stars, last pushed today), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,871 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 skills, from other repositories
webui-code
Use this skill when generating or reviewing WebUI code. This skill contains knowledge on WebUI code following Chrome's Lit-element architecture and styling guidelines.
figma-to-webui
Recreate Figma design frames as production-level Chromium WebUI code following Chrome's Lit-element architecture and styling guidelines. Use when asked to create WebUI code from a Figma frame URL.
browser-agent
Build and deploy a Browser Agent on Omnara, a long-running agent that signs in to web apps with stored credentials and works in a real browser, overnight if needed. Interviews the user about the job, sign-ins, schedule, and Slack or Discord first. Use when the user asks to build, deploy, set up, update, or remove a…
agent-browser
Drive a real browser with the agent-browser CLI (headless Chromium over CDP). Use for any browser task: verifying a UI change, taking a screenshot, walking a web page, filling a form, extracting page text, reproducing a frontend bug, or attaching UI evidence to a PR, issue, Linear comment, or Slack thread. Starts with…
webcrawler-deep-crawl
Deep-crawl any website from start URLs, return per-page LLM-ready text/markdown/HTML plus metadata (title, description, author, language, canonical URL, OG) and in-scope outbound links. Use when user mentions deep crawl website, recursive crawl, crawl a whole site, scrape entire website, scrape docs site, scrape…
ebay-search-listing
Extracts product listings from any eBay search or category page URL, returning per-item cards (itemNumber, url, title, subtitle, caption, price, priceWithCurrency, currency, wasPrice, bids, shipping, seller, sellerFeedbackCount, sellerPositiveRating, reviewsCount, starRating, image) plus pagination state (currentPage…