Performance Hub LogoPerformance Hub

Skills

Skills are packaged know-how — repeatable procedures the agent has learned, kept, and can run on demand. Where memory holds facts, a skill holds a method: how to produce the weekly report, how to format the newsletter, how to reconcile the till export.

Your agent arrives with a set of built-in skills that make it substantially more capable than a stock AI assistant, and it can learn new ones — from you, simply by being shown or told.

The built-in skills

Every AI Agent Processor ships with skills maintained by Performance Hub and kept current with software updates. The headline ones:

The operational runbook (ph-runbook)

The quiet superpower. The runbook makes the agent an expert in its own situation: it knows it's an appliance at your facility, what's connected to what, how its Performance Hub access works, where its own files and settings live, and how to get things done in this specific environment rather than in the abstract.

You don't invoke it — you just notice its effects. It's the difference between an assistant that says "you could try checking your access control system" and one that says "two door controllers at this facility haven't reported in over an hour — want the details?". It's also why the agent can sensibly answer questions about itself — its backups, its spend, its connection — and knows the right way to act on your operational data. The runbook keeps the agent effective as the product evolves, without you re-briefing it.

Notifications (ph-notify)

Lets the agent reach you proactively — real push notifications to your phone and browser when something you asked it to watch for happens. Covered fully in Notifications.

Migration (ph-agent-migration)

Brings an existing AI assistant's life — chat history, memory, files, skills, schedules — across from another computer or platform. Covered in Migrating to your agent.

Onboarding

Runs the first-run conversation that introduces the agent and captures your initial context — see Getting started.

Built-in skills are part of the product: they update with the software, and you don't manage them.

Teaching your agent new skills

The interesting part. When the agent does something well that you'll want again, make it a skill — conversationally:

"That's exactly the format I want for the weekly report. Save that as a skill and use it every time I ask for the weekly report."

Or teach one from scratch by describing the procedure:

"Learn this as a skill called 'end of month': pull the month's attendance and revenue figures, compare to the previous month, write a one-page summary in the workspace, and notify me when it's ready."

The agent writes the procedure down as a skill, confirms its understanding, and from then on "run end of month" does the whole thing. Skills you create can be:

  • Refined — "update the weekly report skill: add a churn section".
  • Listed and inspected — "what skills do you have?", "show me what the end-of-month skill does".
  • Removed — "delete the old newsletter skill".
  • Combined with automations — a skill defines how, a schedule defines when. The pair is how you build genuinely unattended workflows.

The craft is the same as delegating to a person: be explicit about the output you want, run it once and correct it, then let it run. A skill that produced the right thing three times will keep producing it.

Skills persist — and are protected

Skills your agent learns are stored on the device, survive software updates, and are included in hourly backups — so a device replacement or restore carries your custom skills along with everything else. They also travel in a migration.