The Notes System

Notes are how your agent remembers, takes direction, and tracks work across conversations. They’re also how you and the agent together teach it the specifics of your site — you don’t have to know the answer in advance. This page explains the three kinds and how you use them.

Your site keeps a set of notes — short, persistent records that live on the site itself, not inside any single conversation. Because they persist, they’re how instructions, accumulated know-how, and open work items carry over from one session to the next, and from one agent or person to another. You can read and write them through your agent.

Notes are also what makes the agent adaptable to your site specifically. Out of the box, an agent understands standard WordPress and WooCommerce. But most real sites have their own particulars — custom fields, a bespoke product configuration, a special taxonomy, a plugin with its own way of doing things. Notes are how you bridge that gap: the agent learns how your setup works, and from then on handles your site correctly instead of guessing. This is what turns a general-purpose agent into one that knows your store.

There are three types, each with a different job.

A word on authorship: the agent can create any of these, not just you. Sometimes you’ll write one yourself; more often you’ll ask the agent to write it — either by saying so directly (“my site handles shipping classes this way, so always check them first, and save that as knowledge”) or through a standing instruction that has it create notes on its own. Who physically writes a note matters far less than what the note is for, so that’s how the rest of this page is organized.

Operator notes — standing instructions

An operator note is direction the agent follows on your behalf. It covers two things: how to behave — “always save new posts as drafts for me to review,” “use a warm, plain tone in product descriptions” — and how your site works — “our products use a custom field called vintage_year; always set it when creating a wine,” “the ‘Featured’ collection is a category, not a tag, so add featured items there.” The agent loads your active operator notes at the start of every session and follows them. Change a note, and the next conversation reflects it — no reconnecting, no settings to hunt through. These are the primary way you both shape how your agent works and teach it the specifics of your setup.

Knowledge notes — reusable know-how

When you and the agent work out something non-obvious — the right way to populate one of your custom fields, a workaround for a quirk in your data, the correct sequence for a multi-step setup on your site — the agent can record it as a knowledge note so future sessions start from the solution instead of rediscovering it. Knowledge carries an approval step: a knowledge note is proposed first and only becomes active guidance once you approve it. That separates who can write something down (anyone, on instruction) from what becomes lasting guidance (your decision), keeping the agent’s accumulated understanding of your site under your control rather than letting it teach itself unchecked.

Between them, operator notes and agent knowledge are how your site’s particulars become part of what the agent reliably knows. You can teach it directly through an operator note, or it can record what it learns through a knowledge note you approve. Either way, the next session already knows how your site works — with no need to re-explain it each time.

Working it out together, then saving it

Here’s the part that makes this genuinely different from configuring a tool. You don’t have to know in advance how to instruct the agent. You figure it out with it, in an ordinary conversation, and then save the result. Collaboration with your agent is the key.

For example, say you have a custom post type and you’re not sure how it’s wired up. You don’t need to be. You can simply say:

“Look at how my Listings post type is structured, work out the steps to create a new one correctly, and once we’ve got it right, save that as a note so you’ll always do it this way.”

The agent inspects the setup, proposes a process, you try it together and refine it until it works — and then it records the finished workflow as a note. The next time, it just follows it. What was half an hour of figuring-out becomes a one-line request forever after.

This works because the agent can write notes too, not just you. So a note is often not something you author from a blank page — it’s the distilled result of a session, captured at the moment you’ve cracked something. You bring the goal and the judgment; the agent does the investigating and the writing. Nobody gets stuck staring at an empty box wondering what the right instruction is.

It’s the natural rhythm of working with the agent: do it together once, save what worked, never re-explain it.

Agent notes — trackable work items

An agent note is the system’s built-in way to track a piece of work. Each note carries a status that moves through a defined lifecycle — open, acknowledged, in verification, resolved — so an item can be opened, worked, checked, and closed rather than just mentioned and forgotten. Notes can be assigned to a particular agent, prioritized (low through critical), categorized (bug, optimization, task, and so on), and discussed through a running comment thread attached to each one.

In other words, it’s a lightweight issue tracker built into your site — think of a task board your agent can read and write the same way you can. It’s deliberately flexible. The same mechanism covers an agent reporting a problem it hit, you logging a task for the agent to do later, assigning work to a specific agent, or capturing a running list of things to follow up on.

Combining them: the real power

The three types aren’t separate features — they work together. The connecting idea: an operator note shapes behaviour, and creating and tracking agent notes is a behaviour. So an operator note can put the work-tracking system on autopilot. A few examples:

• “Whenever you hit a slow or awkward step while working, log it as an agent note tagged optimization.” — your agent quietly builds a backlog of things to improve, without you asking each time.
• “If you can’t finish a task, open an agent note describing what’s left and assign it to me.” — nothing gets silently dropped between sessions.
• “After any bulk update, create an agent note summarizing what changed.” — you get a durable record of large operations, written automatically.

You set the policy once as an operator note; the agent applies it from then on. That’s how a handful of note types turn into workflows tailored to how you actually run your site.

One boundary worth knowing: an agent can create and progress its own work items freely, but it cannot approve its own knowledge — promoting a knowledge note to active guidance always requires your sign-off. The agent proposes; you decide what becomes lasting instruction.

In summary: notes are persistent records that outlive a single conversation. Operator notes are standing instructions and the way the agent learns how your site works; agent knowledge is reusable know-how that you approve; agent notes are trackable work items. Because the agent can write notes too, many of them come out of simply working through a task together and saving what worked — you’re never starting from a blank page. And because operator notes shape behaviour, you can use them to drive the work-tracking system automatically — turning three simple building blocks into an agent that fits your site exactly.