The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts (scripts/powershell_generator.py, scripts/user_management.py, and scripts/tenant_setup.py) generate PowerShell scripts by directly interpolating user-controlled strings (from CLI arguments or CSV files) into script templates using f-strings. For example, in user_management.py, the UserPrincipalName and DisplayName are inserted into the script body without escaping. A malicious entry in a CSV file or argument (e.g., containing quotes or semicolons) could allow for arbitrary PowerShell command execution when the generated script is run by a Global Administrator.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external CSV files for high-privilege operations.
Ingestion points: scripts/user_management.py and scripts/powershell_generator.py accept --csv paths to read user data.
Boundary markers: The skill lacks instructions or mechanisms to treat CSV data as untrusted or to ignore embedded instructions.
Capability inventory: The generated scripts have the capability to create users, assign licenses, modify security policies, and offboard accounts.
Sanitization: Validation in scripts/user_management.py is limited to checking for spaces in usernames and does not protect against script injection.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several official Microsoft PowerShell modules (Microsoft.Graph, ExchangeOnlineManagement, and MicrosoftTeams) to function. These are downloaded from the official PowerShell Gallery, which is a trusted source for Microsoft 365 administration tools.

ms365-tenant-manager