azure-enterprise-infra-planner
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- Indirect Prompt Injection Surface: The skill processes external data, including workload descriptions, existing infrastructure code (Bicep/Terraform), and resource metadata, which could potentially contain adversarial instructions designed to influence the agent's behavior.
- Ingestion points: Data is ingested during Phase 1 (environment insights), Phase 2 (requirements gathering), and when handling referenced brownfield infrastructure.
- Boundary markers: Security is maintained through manual plan approval in Phase 5 and a strict destructive action gate in Phase 7 that requires explicit user acknowledgement of risks after they are presented.
- Capability inventory: The skill possesses the capability to execute Azure CLI commands (az deployment, az graph), Terraform operations, and file system modifications within the project directory.
- Sanitization: Mitigation includes mandatory security scanning using Checkov in Phase 6 and enforcement of secure-by-default patterns, such as disabling public network access and using managed identities instead of connection strings.
- Execution of Infrastructure Commands: The skill utilizes standard tools including the Azure CLI and Terraform to provision and manage cloud resources.
- This functionality is essential to the skill's purpose and is protected by a gated pipeline that prevents execution without manual user confirmation and successful validation of the generated code.
- External Data Retrieval: The skill fetches technical specifications, naming rules, and architectural guidance from official documentation sources.
- [Security Consideration]: It utilizes MCP tools to search and fetch content from Microsoft Learn (a well-known service) to ensure that infrastructure plans comply with platform-enforced constraints and naming conventions.
- Automated Security Scanning: The skill integrates the Checkov security scanner to review generated Infrastructure-as-Code (IaC) files.
- This provides an automated layer of defense by identifying high and critical security misconfigurations in Bicep or Terraform files before any deployment is attempted.
Audit Metadata