The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a custom CLI wrapper unifi-api to execute shell commands for fetching network data. It also frequently uses jq for data processing and op (1Password) for credential management.
[CREDENTIALS_UNSAFE]: While the skill uses 1Password for secret management (a safe practice), it explicitly identifies that certain UniFi endpoints (e.g., /rest/setting, /stat/device) return unmasked secrets such as PSKs, UTM tokens, and syslog keys. It provides instructions to the agent to filter these, but the raw access exists.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled strings from the local network, such as device names, hostnames, and SSIDs. The documentation explicitly warns the agent to treat these as data and not instructions, which is a good mitigation, but the attack surface exists.
[SAFE]: The skill implements several best practices, including enforcing read-only access (verified by permissions), restricting operations to a local LAN (192.168.1.1), and instructing the agent to never pin certificates automatically.

unifi-network