Skip to main content
Everything your assistant learns is inspectable. In the Brain section of your assistant’s page you’ll find Guidance (the playbooks) and Functions (the skills), alongside Knowledge, Contacts, and Transcripts. Each assistant has its own — select the assistant on the left, then open the pane.

​
Guidance: the playbooks

The Guidance pane is a document library — “playbooks that shape how your teammate behaves,” as its own help puts it. Entries are listed on the left, grouped by date; select one and the full playbook renders as a readable document on the right (drag the divider if either side needs more room). Each entry shows its title, tags, and a scope badge — personal, or the team it’s shared with — and the search box and scope/tag filters keep large libraries navigable. A Copy button grabs any playbook’s text for use elsewhere. What you’ll find there:
  • Rules you’ve taught — “Exclude internal transfer rows and net refunds against spend when computing monthly spend.”
  • Procedures you’ve walked through — the quarterly-report steps you explained on a call, including the tricky part about using the exchange rate from the first business day of the quarter.
  • Pitfalls it learned the hard way — when an obvious approach failed and a correction fixed it, the corrected approach gets written down.
  • Built-in playbooks — a set of platform-provided how-tos (marked built-in · read-only) covering common ground like working with spreadsheets and documents. Your assistant can’t change these, but it can create its own tailored variant for you.

​
Functions: the skills

The Functions pane is a catalog of everything your assistant can execute — “learned Python skills and platform primitives — signatures, docstrings, and source in one place.” It’s a searchable card grid, with each card showing the function’s name, a one-line signature, and a short description. Filter by kind:
  • Learned — skills your assistant built while working for you: the monthly-spend pipeline, the report generator, the data cleanup routine. These are the compounding payoff of working together.
  • Primitives — the platform’s built-in capabilities (sending messages, searching contacts, and so on). These are the raw materials; they’re listed so you can see the full toolbox. Opening one notes that its implementation lives in the platform itself rather than as stored code.
Click any card for the detail drawer: the signature, a plain-English summary, a parameters table with types and descriptions, what it returns, what it depends on, and — if you’re curious — the actual implementation, with a Copy button when you want the source elsewhere. A verified badge means the skill checks its own results when it runs, and a linked playbooks count shows how many guidance entries put it to use.

​
They work as a pair

Playbooks and skills reference each other — a guidance entry shows how many linked functions it uses, and a function shows its linked playbooks. That’s the shape of real learning here: when your assistant learns a workflow, it typically saves the skill (the reusable steps as code) and the playbook (when and how to apply it, and what to watch out for) together.

​
Reviewing and changing what’s learned

The panes are for browsing; changes happen in conversation, like everything else:
  • Ask what it learned — “what did you take away from the expenses work?” or “show me the playbooks you have about invoicing.”
  • Correct an entry — “that rule about refunds is wrong — refunds should only be netted within the same month.” It updates the guidance.
  • Remove something — “forget the workflow for the old CRM, we’ve migrated.” It deletes the stale entries.

​
Personal vs. shared

By default, learning is personal — private to the assistant that learned it. Team-wide material — a shared style guide, an agreed SOP, rules every assistant should follow — can be stored in a team’s shared pool, where every current member benefits. If something should apply to the whole team rather than one assistant, just say so when you teach it.