The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes data retrieved from external Azure APIs (Cost Management, Monitor, Resource Graph) and CLI tool outputs (kubectl, azqr) which is then used to generate reports and commands. While typical for management tools, this represents an attack surface where maliciously crafted resource metadata (e.g., resource tags or names) could theoretically influence agent behavior.
Ingestion points: Data enters the context via az rest responses in cost-query/workflow.md, kubectl top outputs in cost-optimization/azure-aks-anomalies.md, and Azure Resource Graph queries in cost-optimization/azure-resource-graph.md.
Boundary markers: The skill does not explicitly define custom delimiters for external data ingestion within its instructions.
Capability inventory: The skill utilizes command execution via az, kubectl, and azqr, and performs file operations such as create_file and Remove-Item in cost-optimization/report-template.md.
Sanitization: No specific sanitization logic for external resource metadata is described in the instruction set beyond standard agent processing.

azure-cost