The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill documentation and output templates include standard Nginx administrative commands (e.g., systemctl reload nginx, nginx -t, apt install nginx). These are necessary and appropriate for the skill's primary purpose of infrastructure configuration and do not exceed necessary privileges.
[EXTERNAL_DOWNLOADS] (SAFE): Installation guides reference official Nginx sources (e.g., nginx.org) for GPG signing keys and source code tarballs. These are reputable sources required for the software setup process.
[PROMPT_INJECTION] (SAFE): No malicious instructions, bypass markers, or role-play patterns (e.g., DAN, jailbreak) were found in the metadata or configuration files.
[DATA_EXFILTRATION] (SAFE): No hardcoded secrets, credentials, or exfiltration patterns were identified. Access to system paths like /etc/nginx/ is strictly for legitimate configuration management.
[SAFE] (SAFE): Category 8 (Indirect Prompt Injection): The skill presents an injection surface via configuration generation. 1. Ingestion points: User infrastructure requirements. 2. Boundary markers: Agent-level delimiters. 3. Capability inventory: Nginx config and shell script generation. 4. Sanitization: Skill provides hardened templates; final validation is the agent's responsibility.

configuring-nginx