The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts scripts/deploy_webapp.py, scripts/resource_status.py, and scripts/cost_analyzer.py utilize subprocess.run with shell=True to execute Azure CLI commands. User-supplied arguments such as --name, --resource-group, and --runtime are directly interpolated into these shell strings without validation or escaping, creating a command injection vulnerability.
[DYNAMIC_EXECUTION]: The management scripts generate and execute shell command strings dynamically based on runtime parameters. This practice, combined with the lack of input sanitization, exposes the host system to arbitrary command execution if an attacker provides crafted resource names or other parameters.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external data.
Ingestion points: Log data fetched from Azure services in scripts/resource_status.py via az webapp log show is processed and displayed by the agent.
Boundary markers: Absent. The script provides no delimitation or warnings to the agent when processing these logs.
Capability inventory: The skill includes several scripts that execute shell commands with the identity of the logged-in Azure user.
Sanitization: Absent. There is no evidence of log scrubbing or escaping before the data is presented to the agent.

azure-expert