Create Generic get_viewer_permission Tool

To safely perform restricted actions like assigning issues, merging pull requests, or adding certain labels, an agent must first know its permission level for the repository. Creating a specific permission-checking tool for each action (e.g., can_assign_issues, can_merge_prs) is not scalable.

A much better approach is to create a single, generic tool that reports the user's permission level, allowing the agent to make an informed decision about its capabilities.

Proposed Solution

Create a new, read-only tool named get_viewer_permission.

Implementation:

1. The tool will use a new GraphQL query to fetch the viewerPermission field on the repository object.
2. This field returns the permission level of the user whose token is being used for the API request.

Output:

The tool will return a single string representing the user's permission level. The possible values are:

• ADMIN: Can read, write, and manage the repository.
• MAINTAIN: Can read, write, and manage issues and pull requests.
• WRITE: Can read and write to the repository.
• TRIAGE: Can manage issues and pull requests.
• READ: Can read the repository.

Usage:

An agent will call this tool once to determine its capability level. It can then use this information to decide whether to proceed with restricted actions. For example, to assign an issue, the agent would first call get_viewer_permission and only proceed to call assign_issue if the returned value is ADMIN, MAINTAIN, or WRITE.