Connecting miinideck to Claude gives it the tools: it can create a page, update one, list what you’ve published. What the connector can’t do is tell Claude when any of that is the right move. So you end up as the prompt — Claude builds a report, stops, and waits, and you type “now share that” for the fortieth time. Worse, on the occasions it does reach for the tool unprompted, it doesn’t know the boundaries: a large page gets pushed through the chat channel and arrives truncated, and a revision gets published as a brand-new link when the person you sent it to is still holding the old one. The Skill is a single file you upload once that closes that gap. It doesn’t replace the connector — it sits on top of it and supplies the judgement.
Try it now
Drop a file — get a private link in seconds. No sign-up.
Drop an HTML file or ZIP bundle, or click to choose.
Single file or ZIP. Max 3 MB.
Up to 3 MB, link self-destructs after 7 days. Sign up free to keep links forever, password-protect them, and store more.
Connect miinideck to Claude first — that’s the connector URL you paste in once, and it’s what actually provides the publishing tools. The Skill is the second, optional half: a file you upload to Claude that describes when publishing is worth offering, which tool to reach for, and where the edges are. Without the connector the Skill has nothing to call, which is the one thing worth getting right about the order.
After that, the behaviour changes rather than the capability. Claude finishes a standalone page — a report, a one-pager, a deck, a dashboard — and offers a private link instead of waiting to be asked. It doesn’t publish silently: it builds the thing, offers, and acts once you say yes. If you already asked for a link, it just does it, and it can set a password, an expiry or a project on the way out rather than making you go and configure them afterwards.
The part that saves the most trouble is knowing when not to use it. Anything Claude passes through a tool call is text it has to generate, so a large page — roughly past a hundred kilobytes — would arrive cut off. The Skill tells it to stop and point you at the uploader or the CLI, where the file’s actual bytes are sent, instead of silently truncating your page. It also tells Claude to revise with the update tool rather than the create tool, so a second draft keeps the address you already handed out.
The connector is enough to publish. The Skill is worth adding if you publish often, because it changes who does the remembering — Claude offers at the right moment, keeps the same link on revisions, and refuses to force a large file through the chat channel. If you publish once a month and are happy to ask each time, the connector on its own is fine. They aren’t alternatives: the Skill calls the connector’s tools, so it needs the connector present.
A file you upload to Claude once that gives it instructions for a particular kind of task — in this case, publishing a finished page as a private link. It sits alongside your conversations rather than inside any one of them, so the behaviour is there next time without you setting anything up again. Nothing installs on your machine, and it can be removed as easily as it was added.
No — the Skill is explicit that it shouldn’t. It builds the page, then offers, and waits for a yes. The exception is the obvious one: if you already asked for a link, it treats that as the yes rather than asking again. Given that a published link is a real page at a real address, that ordering is deliberate.
It tells you to use the uploader or the CLI instead, and that’s the correct answer rather than a limitation of the hosting. A tool call carries the document as text Claude generates, so it can only be as large as a single response — past roughly a hundred kilobytes it would be truncated. Uploading directly sends the file’s bytes, which is why a large single page or a whole multi-file build goes that way. The storage side handles far bigger than the chat channel does.
Not if the Skill is doing its job — it tells Claude to revise with the update tool, which keeps the same address. That’s the difference between sending someone a link and sending them a link they can keep: the client who bookmarked the first draft opens it a week later and sees the current one rather than a page that no longer matches what you talked about.
Claude Code has its own path — it can publish over the connector or the CLI straight from the terminal, which is covered on its own page. The Skill is aimed at the chat surface, where you’re building a page in conversation rather than in a project directory. Nothing stops you having both connected.
Yes, by default. The address is a 32-character unguessable string, the page carries a noindex tag, and there’s no public directory here to browse. A link can still be forwarded by whoever you send it to, which is why the Skill can set a password or an expiry at the point of publishing — worth doing at the moment you create the page rather than deciding later.
No card to try, no sign-up to get a link. Sign up free to keep links forever, password-protect them, and store more.