The Agent Skills Directory

[SAFE]: The skill is designed for VPN administration and diagnostic tasks. All included commands (e.g., docker, nft, ip, systemctl) are standard for the management of network services and containers.
[DATA_EXPOSURE]: While the skill interacts with sensitive directories such as /opt/amnezia and configuration files like awg0.conf, it includes proactive security guidelines. It specifically instructs: 'Do not log PrivateKey, preshared key, obfuscation secrets entirely. In answers — placeholders.' and 'Archive with keys — only on server/secret media, not in wiki.'
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests untrusted data from docker logs and journalctl. However, this is a standard requirement for diagnostic skills. The risk is mitigated by explicit instructions to redact sensitive information and the fact that the ingested data is local system output rather than attacker-controlled remote content.
[COMMAND_EXECUTION]: The skill uses administrative commands for firewall management (ufw, nft, firewall-cmd) and service control. These are used appropriately within the context of VPN setup and do not exhibit patterns of unauthorized privilege escalation or persistence beyond the intended service operation.

amnezia-vpn