The Agent Skills Directory

[COMMAND_EXECUTION]: Uses the standard az and jq binaries to interact with Azure cloud services, which are appropriate for the skill's stated management purpose.
[CREDENTIALS_UNSAFE]: Includes a command to fetch Azure storage account keys (az storage account keys list). While functional for administration, this action exposes sensitive, long-lived credentials to the agent's context.
[PROMPT_INJECTION]: The skill processes external resource names, creating an indirect prompt injection surface. Ingestion points: Container and blob names are ingested via listing commands in SKILL.md. Boundary markers: No delimiters or instructions are present to prevent the agent from interpreting names as commands. Capability inventory: The skill can perform file uploads, downloads, and deletions using these names. Sanitization: No sanitization or validation of the cloud metadata is performed before the agent processes it.

azure-storage