The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from the Azure environment, such as resource names, tags, and configuration properties, which could contain malicious instructions.
Ingestion points: Data is fetched via MCP tools and the az CLI as described in SKILL.md (Step 2: Resource Discovery & Analysis).
Boundary markers: The skill instructions do not specify the use of delimiters or provide warnings to the agent to disregard instructions embedded within the resource metadata.
Capability inventory: The skill has the capability to write markdown files to the local workspace and execute commands in the terminal (SKILL.md).
Sanitization: There are no instructions for sanitizing or escaping the data retrieved from Azure before it is interpolated into the generated architecture documentation.
[COMMAND_EXECUTION]: The skill utilizes the az CLI to perform resource discovery and detailed configuration analysis.
Evidence: SKILL.md explicitly instructs the agent to use terminal commands such as az resource list and az network vnet show for data retrieval.
Context: While these commands are intended for read-only analysis, the execution of shell commands using identifiers retrieved from the environment creates a potential surface for command injection if resource names or variables are not handled securely.

azure-resource-visualizer