The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches installation scripts and configuration files from external repositories and services, including GitHub (NousResearch), Cloudflare, and Cloudsmith.
[REMOTE_CODE_EXECUTION]: Executes the Hermes Agent installation script by piping content from GitHub directly to the bash shell.
[COMMAND_EXECUTION]: Performs numerous shell operations via SSH on the target VPS to install software, move files, and manage system resources.
[PRIVILEGE_ESCALATION]: Utilizes sudo and root privileges to install system packages via apt-get, configure web servers (Caddy), and register systemd services.
[PERSISTENCE]: Modifies shell profile files (~/.bashrc and ~/.profile) to manage environment variables and installs four systemd services (vps-gate, vps-tunnel, hermes-gateway, hermes-dashboard) to maintain always-on operation.
[DATA_EXFILTRATION]: The sign-in gate service (gate.py) transmits visitor IP addresses to an external geolocation and metadata API (ip-api.com) for network verification checks.
[INDIRECT_PROMPT_INJECTION]: The skill architecture creates a surface for indirect prompt injection by processing untrusted data from external sources like Telegram or Discord chat messages and user-created dashboard modules.
Ingestion points: User messages from chat gateways and HTTP POST data from the dashboard interface.
Boundary markers: Uses SOUL.md and CLAUDE.md files to define agent rules and behavior constraints.
Capability inventory: Access to shell command execution, filesystem modification, and network operations on the remote server.
Sanitization: Implements HTML escaping and URL encoding in web responses, and uses bcrypt for password hashing.

setup-vps