dashieslisted
Install: claude install-skill Dashies-ai/dashies-plugin
# Building a dashboard on Dashies
**Dashies keeps a dashboard alive.** You author it once by writing a **spec** - one small YAML
document - and pass it to `publish_dashboard`'s `spec` argument. The server turns that into the
dashboard, checks it, runs your SQL once to put real numbers in it, and then re-runs that SQL on
the schedule you chose, with no AI in the loop. The result is a URL whose numbers stay current
and which costs nothing per view.
**Four things decide what a dashboard SAYS, and every spec needs all four:**
1. **which connection** the numbers come from,
2. **one read-only `SELECT` per dataset**,
3. **what each of its columns means** - which are the things people group and filter by, and
which are the numbers,
4. **how often it should refresh.**
Everything under that - how the data is prepared, and where it is kept - is the server's to
decide. It decides it from the four things above and nothing else, so the same document against
the same connection gets the same answer and a republish that changes nothing changes nothing.
**What the connection can do is one of those inputs**, so if it gains a capability the answer may
legitimately move - which is why the publish report tells you what it decided every time rather
than only when you changed something.
**You do not choose it and you do not need a vocabulary for it.**
**What the page is BUILT OUT OF is a separate question, and the answer is the USER'S.** Give the
server a `tiles` layout and it draws the pag