> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cruq.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Skills & schedules

> Reusable procedures an agent loads on demand, and recurring unattended runs.

## Skills

A skill is a named procedure - a block of instructions the agent can look up
mid-run rather than carrying in its system prompt.

The agent gets two builtins: one to **list** the skills available to it, and one
to **load** a skill by name. It sees the names and descriptions up front and
pulls in a full body only when it decides the task calls for it.

That is the point: ten long procedures pasted into a system prompt cost tokens
on every single turn and dilute the model's attention. As skills, they cost a
one-line summary each until they are actually needed.

Skills are workspace-scoped and managed under **Knowledge → Skills**. Good
candidates are procedures with steps and edge cases - a refund policy, an
escalation path, a release checklist - rather than facts, which belong in
[Knowledge](/docs/concepts/knowledge).

## Schedules

A schedule runs an agent unattended on a recurring basis with a fixed prompt -
the daily brief, the weekly digest, the hourly sweep.

Set a frequency (hourly, daily, or weekly), a time and timezone, and the input
to send. Each firing is an ordinary run: fully traced, metered, and visible in
AI Monitoring.

<Note>
  Scheduled runs count against your plan's monthly run quota, and the quota is
  checked before each firing. A workspace over its limit stops firing rather than
  accruing overage. Plans also cap how many schedules can be enabled at once - see
  [Plans & limits](/docs/plans).
</Note>

Schedules are managed per agent, in the agent's **Schedules** section.