entra-id-auth
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- Deceptive Metadata and Knowledge Override: The skill documents measurements and tool versions (e.g., Azure CLI 2.90.0) using dates from the year 2026. This is a security consideration as it may attempt to override the agent's current, verified knowledge with unverified future-dated information, potentially leading to incorrect configuration or the use of non-standard command flags based on erroneous version measurements.
- SQL Injection Risk in Database Templates: The provided T-SQL scripts for user creation use string concatenation to build commands (e.g., building a command using + @principal_name +). This represents a potential vulnerability surface for indirect prompt injection or SQL injection if the principal name is sourced from untrusted user input without adequate sanitization or the use of safe functions like QUOTENAME.
- Dynamic Execution and Command Patterns: The instructions utilize dynamic SQL execution (EXEC (@cmd)) and local container diagnostics (docker exec, docker inspect). While standard for administrative tasks, these involve the assembly and execution of commands at runtime, which requires careful validation of parameters to mitigate command and SQL injection risks.
Audit Metadata