---
title: "Skills & memory"
description: "Reusable instructions, remembered facts and settled project decisions — given to every agent run PaneScale starts, always as context and never as a command."
url: "https://panescale.com/docs/skills"
updated: "2026-10-13"
---

# Skills & memory

Reusable instructions, remembered facts and settled project decisions — given to every agent run PaneScale starts, always as context and never as a command.

Agents start every run without knowing your conventions, what you told them last week or what the team already decided. **Settings → Skills** keeps that knowledge in one place and hands it to every agent run PaneScale starts for you.

| Kind | Holds | Example |
| --- | --- | --- |
| **Skills** | Longer instructions you reuse | A review checklist, code conventions, your release steps |
| **Memory** | Short facts — for all projects or for one | *Answer in English* · *Tests run with `npm run test:all`* |
| **Decisions** | Settled decisions of a project, so agents don’t reopen them | *We stay with SQLite instead of Postgres* |

## Where they are used

Skills, memory and decisions go into every agent run that PaneScale starts on your behalf: [Chat](https://panescale.com/docs/chat) and project threads, [teammates](https://panescale.com/docs/agent), [tasks](https://panescale.com/docs/tasks), [routines](https://panescale.com/docs/routines), the [Overnight Autopilot](https://panescale.com/docs/autopilot), [Point & Fix](https://panescale.com/docs/point-and-fix) and the [Video Studio](https://panescale.com/docs/video-studio) agent.

Interactive terminal panes are your own sessions — nothing is slipped into them.

> **Context, never commands**
>
> Everything is added to the prompt clearly marked as context. A skill or a remembered fact can inform an agent, but it can’t override the rules of a run — for example that the autopilot never pushes.

## Skills

**Add skill** with a name and the content — *What should the agent always keep in mind?* **Edit** changes a skill, the bin deletes it. Skills apply to all projects; up to ten are given to each run, in alphabetical order.

## Memory

Type a fact and press `Enter` or **Add**:

- **Fact for all projects** — applies everywhere,
- **Fact only for** the active project — appears when a project is open.

Every fact shows whether it applies to all projects or one, and when it was added. PaneScale keeps at most 50 facts per project; the newest 20 that apply go into every run.

## Decisions

Decisions belong to the active project — open a project to record them. Up to 50 are kept per project, and the newest 20 go into every run. Use them for things that were weighed once and shouldn’t be discussed again with every agent.

## Teammate memory

[Teammates](https://panescale.com/docs/agent#memory) also have a memory of their own, which they can learn from their conversations. It works the same way — context, never instructions — and only applies to that teammate.