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

# Skills

> Reusable capabilities the agent runs to do real work — built in, or authored by your team.

**Skills** are packaged, reusable capabilities the agent can run — scripts and
procedures for concrete tasks (for example, setting up cloud IAM). Rather than
improvising every time, the agent reaches for the right skill, which makes its
work more reliable and repeatable.

## How the agent uses skills

When a task matches a skill, the agent loads and runs it as part of a session —
you'll see it in the tool activity. Skills run with the credentials you've
[selected for the session](/agent/sessions), under the same guardrails as any
other agent action: changes still surface as [plans](/agent/plans) for approval.

<Note>
  The set of built-in skills evolves as the product grows. You don't install
  built-in skills manually — the agent draws on them automatically.
</Note>

## Your team's own skills

Built-in skills are not the whole set. A team can author its own under **Skills**
in the sidebar — the runbook you would otherwise paste into every session, stored
once and matched like any built-in. The agent can also write one for you at the
end of a session, when you have just walked it through something it should
remember how to do.

Team skills obey the same rules as built-in ones: they run with the credentials
selected for the session, and anything they change still surfaces as a
[plan](/agent/plans).

## Working alongside your terminal agent

If you use **Claude Code** or **Codex**, you can attach one of their sessions to a
Nuphos conversation from the **+** menu in the composer, so the agent can read
what happened there. In the other direction, Nuphos exposes an
[MCP server](/connectors/mcp) that lets an external agent query your
infrastructure and open resources in Nuphos.