The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands, specifically az (Azure CLI) and dot (Graphviz), to query cloud environments and generate architectural diagrams.
[CREDENTIALS_UNSAFE]: The skill instructions include commands like az webapp config appsettings list to trace dependencies. These commands retrieve environment variables that often contain sensitive plain-text credentials such as API keys and connection strings, exposing them to the agent's context.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from Azure resource names, tags, and properties. This external data could potentially contain malicious content intended to influence the agent's output.
Ingestion points: Data is ingested through JSON output from multiple az commands.
Boundary markers: The instructions do not define specific delimiters to isolate external resource data from the agent's instructions.
Capability inventory: The skill possesses broad shell execution capabilities for interacting with the Azure CLI and local diagramming tools.
Sanitization: There is no mention of sanitizing or validating resource metadata before it is incorporated into reports or diagrams.
[EXTERNAL_DOWNLOADS]: The skill identifies the resource-graph Azure CLI extension as a prerequisite, which is a standard Microsoft-provided extension for enhancing query capabilities.

azure-resource-discovery