The Agent Skills Directory

Category 8: Indirect Prompt Injection (LOW): The skill defines a workflow where untrusted data (CSV files) is processed to generate high-privilege PowerShell scripts.
Ingestion points: scripts/user_management.py and scripts/powershell_generator.py take --csv inputs and user-defined --policy-name strings.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are documented in the prompt templates.
Capability inventory: The resulting scripts are executed with Global Administrator or Security Administrator privileges, capable of modifying all tenant configurations.
Sanitization: Implementation details are missing as the Python logic scripts were not included in the bundle, making it impossible to verify if input is escaped.
Category 10: Dynamic Execution (LOW): The skill's primary function is generating PowerShell code at runtime based on user templates. While this is the intended purpose, it creates a potential execution vector for the agent.
No Code (NO_CODE): The documentation references three Python scripts (powershell_generator.py, user_management.py, and tenant_setup.py) that are not present in the provided files. The analysis is limited to the markdown templates and metadata.
Category 4: External Downloads (SAFE): The skill references official Microsoft modules (Microsoft.Graph, ExchangeOnlineManagement). These are from a trusted source (Microsoft) and are standard for the stated task.

ms365-tenant-manager