azure-reliability
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Infrastructure Command Generation and Execution: The skill generates and executes Azure CLI (
az), Azure Developer CLI (azd), and Terraform commands to assess and modify cloud infrastructure. While the skill includes explicit user confirmation steps, the capability to perform automated, high-privilege operations on production resources is a notable security consideration. - Dynamic Infrastructure-as-Code (IaC) Patching: The skill programmatically modifies local Bicep and Terraform files to update reliability settings such as zone redundancy and storage SKUs. Users should verify these automated patches to ensure they comply with organizational infrastructure standards before applying them.
- Sensitive Data Access: To facilitate storage account migrations, the skill retrieves Azure storage connection strings using
az storage account show-connection-string. This behavior is documented as part of the remediation workflow, but the handling of these credentials during the session is a point for user awareness. - Application Source Code Modification: In specific scenarios, such as configuring health probes for Azure Functions on certain plans, the skill may propose adding HTTP-triggered endpoints directly to the application source code. The skill is designed to request explicit consent before modifying source files.
- Indirect Prompt Injection Surface: The skill ingests metadata from Azure Resource Graph queries, which is then used to construct assessment tables and remediation plans. This introduces a potential surface for indirect prompt injection if resource names or tags contain unexpected instructions.
- Ingestion points: Metadata entering the context via Azure Resource Graph queries in
SKILL.md(Phase 1/2) andreferences/zone-redundancy-checks.md. - Boundary markers: The instructions do not specify the use of delimiters or specific safety warnings when processing ingested resource metadata.
- Capability inventory: Significant capabilities include executing shell commands (
az,azd,terraform), writing to the local file system (IaC patching), and modifying application settings and source code. - Sanitization: The skill uses standard CLI command construction; however, explicit sanitization of ingested resource names or properties before their inclusion in generated commands is not detailed.
Audit Metadata