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

> ## Agent Instructions
> Use the exact commands, endpoints, and product terms in these docs. Commands that start with `pctx` mean the Port of Context platform CLI installed from pctx.ai, not the open-source tool of the same name.

# Skill policies

> Require review before skills are shared, and control external import.

Open **Admin** → **Skills** to set these policies.

| Policy | Effect |
| - | - |
| **Require review before publishing** | New skills stay personal until an admin approves them for the workspace. |
| **Allow external skill import** | Members can copy skills from the public registry and supported repository URLs. |

## Review a skill

1. The author creates the skill as personal and tests it.
2. The author submits it for review. It stays personal while it waits.
3. An admin reads the `SKILL.md`, description, source, and proposed scope.
4. The admin approves it, which publishes it to the workspace, or rejects it, which returns it to the author.

## What to check

Before approving a skill, confirm that it:

* Has a clear trigger and outcome
* Uses only the access it needs
* Contains no credentials or tenant-specific secrets
* Refers to stable context, not one person's local paths
* Handles destructive or expensive actions explicitly

An imported skill is a copy of someone else's instructions and files. Treat it as untrusted until someone has reviewed it.

Access to a skill does not grant access to the services it mentions. Service and tool rules still apply when the skill runs.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.