entra-poc-advisor
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFE
Full Analysis
- [Administrative Tooling Patterns]: The skill is designed to assist with complex cloud identity configurations. It generates PowerShell scripts and Markdown documentation that require manual review and execution by an administrator, maintaining a clear boundary of control.
- [Safety Guardrails and Dry-Run Support]: The skill instructions and generated PowerShell scripts (e.g.,
Deploy-PrivateAccessQuickAccess.ps1) implement safety features such as the-WhatIfparameter. This allows users to preview changes before applying them, which is a standard security best practice for infrastructure automation. - [Operational Mode Constraints]: The skill defines clear operational modes (Guidance Only, Read-Only, Read-Write). This design ensures the agent cannot access or process tenant data through its MCP server tools without explicit user authorization for each mode transition.
- [Indirect Prompt Injection Surface]: The skill reads external data from Microsoft Entra via the Graph API. This creates an attack surface common to all agents processing external content. However, the skill includes explicit instructions to use discovered schemas and not to fabricate data, which mitigates the risk of accidental obedience to instructions embedded in tenant metadata.
Audit Metadata