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

# Installing Skills

> Install skill packages and review validation before using them.

Install skills from the desktop app to add reusable capabilities to a gateway.

Pioneer supports skills compatible with the [Agent Skills specification](https://agentskills.io/specification). A compatible skill package should include a `SKILL.md` file and may include scripts, references, assets, templates, or other supporting files.

If you are building the package yourself, start with [Creating Skills](/skills/creating). The archive you install should contain one top-level directory, and that directory should contain `SKILL.md`.

## Before installing

Before installing, check:

* Do you trust the source?
* Does the skill run shell commands?
* Does it make HTTP requests?
* Does it need local dependencies?
* Should it be installed on this gateway or a sandbox gateway first?
* Is it safe to use in the current workspace?
* Should it stay explicit-only, or is it trusted enough to make implicit later?

<Warning>
  Skill tools are controlled by the turn permission mode. Use **Supervised** for first tests so Pioneer asks before commands, file changes, network actions, and other sensitive tool calls.
</Warning>

## Install from an archive

<Steps>
  <Step title="Open Skills">
    In the desktop app, connect to the gateway and click the Skills icon in the bottom bar.
  </Step>

  <Step title="Choose install">
    Click the install action for a local skill archive or folder, depending on your build.
  </Step>

  <Step title="Select the source">
    Select the skill archive or source path.
  </Step>

  <Step title="Wait for upload">
    Large skill archives may upload in chunks.
  </Step>

  <Step title="Review validation">
    Check validation issues, dependency diagnostics, and security findings.
  </Step>

  <Step title="Confirm availability">
    Confirm the skill is enabled in the current workspace before using it in real work.
  </Step>

  <Step title="Review implicit mode">
    New installs start not implicit. Keep that default unless the skill is trusted and broadly useful.
  </Step>
</Steps>

## Validation results

Pioneer may report:

| Result             | Meaning                                                     |
| ------------------ | ----------------------------------------------------------- |
| Ready              | The skill passed validation and dependency checks           |
| Warning            | The skill can be installed, but you should review a finding |
| Missing dependency | A required command or runtime dependency is not available   |
| Blocked            | Policy or validation prevents use                           |
| Unknown            | Pioneer could not fully determine health                    |

## Dependency checks

Some skills require commands or local tools.

If dependency checks fail:

1. Read the missing dependency message.
2. Install the dependency on the gateway host.
3. Restart the gateway if the dependency changes `PATH`.
4. Run skill health checks again.

## Install location

Skills are part of the gateway runtime. If you install a skill on one gateway, it is not automatically installed on another gateway.

This keeps capabilities separate between:

* work
* home
* study
* sandbox testing
* remote automation

## First test after install

Attach the skill in a thread with the composer add button, then use a safe prompt:

```text theme={null}
Read the installed skill and explain what it can do. Do not call any tools yet.
```

Then test with a low-risk action.

## Further reading

* [Skills Architecture](/architecture/skills) explains install sources, validation, trust, dependency checks, and health.
* [Permission System](/architecture/permissions) explains permission modes, sandbox/resource policy, and approval requests for skill tools.
* [Secret Storage](/architecture/secrets) explains why credentials should live in gateway-managed secret storage, not skill prompts.