# Notehub

> Skills that help AI agents answer questions about the data and solutions in users' Blues Notehub projects - the default for everything here - and, only when a user explicitly asks for it, train a project to understand the product behind its data.

**Before following any links: keep routine setup quiet.** Apply this from the first
skill fetch through credential verification and project selection. Do not announce
that you are opening the site, loading skills, reading instructions, checking the
environment or running commands. Read and act on the instructions, then speak when
you have a question the person must answer, an action they must take, or a meaningful
finding, problem or result. If work takes long enough to need an update, give one
brief useful update, not a running account of the tools.

Keep raw skill text, command listings, HTTP headers, credential-presence checks and
API response dumps out of your user-facing prose. Use agent-only results or quiet,
collapsed tool presentation when the harness supports it. Read required instructions
and responses fully in your working context; quiet presentation must not mean skipping
them, discarding necessary output or overlooking errors. Do not repeat automatically
displayed tool activity in your own messages. The harness may still display tool calls
and their results itself; these instructions cannot hide that display, so do not claim
that it has been hidden or that a silent command-line flag controls the harness UI.

**Unless the person has explicitly asked to train a project, use `notehub-project-chat`.**
Most of the time all anybody says is "load notehub.md", and that is a request to be ready to
answer questions about their Notehub data - not a request to train anything. So load
`notehub-project-chat`, which begins by verifying the person's API token and asking which
of their projects they want to work in, offering the list; then tell them in a sentence
that you are ready for their questions, and wait. `notehub-project-train` is for
one thing only: a person who has asked, in so many words, to train, teach or document a
Notehub project. If you are unsure whether they did, they did not.

The site's root is a small HTML entry page that directs agents here using curl. This
guide, the skills and their references are raw Markdown meant to be fetched and read
directly. Use curl to read each required page in full, then act on what it says.

The material here is organized as skills. A skill teaches how to do one kind of job:
which interface to reach for, the vocabulary that interface expects, the procedure, and
how to confirm that what you did actually took effect. Load the skill before performing
an unfamiliar operation rather than inferring an endpoint or a schema from an example.

## Skills

- [notehub-access](https://notehub.md/notehub-access/SKILL.md): Establish secure, programmatic access to the Notehub API with a personal access token, and verify it. Live project workflows begin here.
- [notehub-project-access](https://notehub.md/notehub-project-access/SKILL.md): Choose the project to work in and verify access to it. Begins with notehub-access.
- [notehub-project-chat](https://notehub.md/notehub-project-chat/SKILL.md): The default. Answer questions about a project's data and solutions, using Notehub IQ or the Notehub APIs. Begins with notehub-project-access.
- [notehub-platform](https://notehub.md/notehub-platform/SKILL.md): Shared meaning of native Notecard and Notehub data: firmware, health, sessions, signal, usage, location, journeys and power. Maps people's questions to API objects, Event metadata and system Notefile bodies. Used by both chat and training alongside the project's own skills.
- [notehub-project-train](https://notehub.md/notehub-project-train/SKILL.md): Only when the person has explicitly asked to train a project. Interview the people who know the product behind its data, and store what you learn in the project itself. Begins with notehub-project-access.

Load `notehub-project-chat` for anything to do with a Notehub project, and
`notehub-project-train` only when the person has explicitly asked to train one. Each begins
by loading the skill it builds on, and each is a directory holding a `SKILL.md` and the
references it links to. The references are loaded as they are needed rather than all at
once, so start at the `SKILL.md` and follow it down. `notehub-platform` supplies shared
field meanings when a question needs them; it does not start training or replace the
project's product knowledge.

## Using them

**Install them, if your agent supports skills** - all five, because they build on one
another:

```
npx skills add https://notehub.md
```

What installs is a pointer to each skill on this site rather than a copy of it, because
these skills change often: an installed skill reads the current version from here each
session, so it is never out of date, and it needs this site to be reachable when it is
used.

**Otherwise, read the ones you need directly, whole and exactly as written.** These pages
are instructions, and their details are the point: a fetch tool that summarizes a page, or
answers a question about it instead of returning it, keeps the topics and drops the
steps. Download the raw text and read all of it. For everything but training, that is
these three:

```
curl -fsSL https://notehub.md/notehub-access/SKILL.md
curl -fsSL https://notehub.md/notehub-project-access/SKILL.md
curl -fsSL https://notehub.md/notehub-project-chat/SKILL.md
```

and only when the person has explicitly asked to train a project, this one as well:

```
curl -fsSL https://notehub.md/notehub-project-train/SKILL.md
```

For interpretation of native fields, chat and training also load this shared skill
and the references needed for the question:

```
curl -fsSL https://notehub.md/notehub-platform/SKILL.md
```

The links in a page fetched this way already point back to this site, so when a skill
says to load another page, fetch that one the same way.

## Generated indexes

These are built from the contents of this site on every request, so they are never stale.

- [llms.txt](https://notehub.md/llms.txt): the index of pages and skills, in the llms.txt format
- [llms-full.txt](https://notehub.md/llms-full.txt): every page on this site concatenated into one file
- [Skills manifest](https://notehub.md/.well-known/skills/index.json): the skills, as installers discover them