The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/agent.py script uses subprocess.run to execute the curl utility for interacting with Cisco ISE APIs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external Cisco ISE API endpoints, creating an indirect prompt injection attack surface.
Ingestion points: API response data is fetched via get_network_devices, get_authorization_policies, and get_active_sessions functions in scripts/agent.py.
Boundary markers: No delimiters or explicit instructions are used to prevent the agent from misinterpreting ingested data as instructions.
Capability inventory: The script can write JSON report files and execute shell commands (curl).
Sanitization: No sanitization or validation logic is applied to the API response content before processing.

implementing-network-access-control-with-cisco-ise