The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the UniFi Network Controller API, including client names, device aliases, and network descriptions, which are ingested into the agent context.
Ingestion points: Data retrieved via various tools such as list_clients, list_devices, and list_networks described in SKILL.md.
Boundary markers: Absent; the skill does not define specific delimiters or instructions for the agent to ignore potentially malicious embedded instructions in the controller data.
Capability inventory: The presence of high-privilege tools like create_firewall_rule, delete_network, restart_device, and download_backup across the management scripts creates a risk if the agent is manipulated by injected data.
Sanitization: No explicit sanitization or validation of data retrieved from the controller is mentioned locally within the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs users to install a third-party package from PyPI and download source code from an external repository.
Package: unifi-mcp-server is required for use.
Repository: References https://github.com/enuno/unifi-mcp-server.git for source installation.
[COMMAND_EXECUTION]: Documentation suggests configuration options that bypass critical security checks for network communication.
Evidence: The Troubleshooting section in SKILL.md suggests setting UNIFI_VERIFY_SSL=false to resolve connection issues. This disables SSL certificate verification, making the administrative traffic and credentials vulnerable to man-in-the-middle (MITM) attacks.

unifi-mcp-server