The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The provided automation script (scripts/agent.py) presents a potential command injection surface on target network switches.
Ingestion points: Variable inputs for vlan_id and vlan_name are used in several functions including create_vlan, configure_access_port, and configure_trunk_port.
Boundary markers: The skill instructions do not provide delimiters or instructions for the agent to treat user-supplied network plans as data rather than instructions.
Capability inventory: The script utilizes the netmiko library to execute privileged configuration commands (send_config_set) on network switches.
Sanitization: The script performs no validation, range checking, or escaping on the vlan_id or vlan_name variables before interpolating them into command strings (e.g., f"vlan {vlan_id}").
[COMMAND_EXECUTION]: The scripts/agent.py script accepts sensitive authentication credentials (--username and --password) as plain-text command-line arguments. This is a common but insecure practice as these credentials may be exposed in system process lists or shell history logs.
[METADATA_POISONING]: There is an inconsistency in the author attribution. The skill frontmatter in SKILL.md lists the author as 'mahipal', whereas the LICENSE file attributes the copyright to 'mukul975'. While this may be a simple oversight, it constitutes misleading metadata.

configuring-network-segmentation-with-vlans