The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The documentation contains a hardcoded Vultr API key ZO5RLHQVKSLJP4QMTUOIMFUGHTIT26FVPZVQ. While presented as a prerequisite example, it follows the format of an active credential.
[DATA_EXFILTRATION]: The skill instructs the agent to read sensitive local files, specifically SSH private and public keys (~/.ssh/id_rsa, ~/.ssh/id_ed25519, ~/.ssh/id_vultr.pub) and environment configuration files (~/.hermes/.env) containing API keys for services like ElevenLabs.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to manage system processes (pkill, pgrep), check network listening states (ss), and manage remote VPS servers via ssh and scp.
[PRIVILEGE_ESCALATION]: The instructions require sudo permissions for administrative tasks such as installing software (apt-get install), modifying system configuration files in /etc/, and managing systemd services.
[PERSISTENCE]: The skill implements persistence by creating and enabling custom systemd unit files (e.g., microsocks.service) on the remote host to ensure proxy services are automatically restarted on system boot.
[REMOTE_CODE_EXECUTION]: The skill executes remote commands on VPS instances via SSH, including installing packages (pip install proxy.py), running background processes with nohup, and executing complex multi-line logic passed through SSH pipes.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs software dependencies from external package registries via pip, npm, and apt. These resources are fetched and executed immediately upon deployment.
[DYNAMIC_EXECUTION]: The skill generates and executes Python and Node.js code at runtime, using python3 -c for inline data processing and dynamically creating script files (e.g., simpleproxy.py and decode.py) for execution on remote hosts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external API responses (Vultr, OpenAI, GitHub) and processes it without explicit sanitization. Evidence chain: (1) Ingestion points: Vultr API responses in SKILL.md and templates/startup.sh; (2) Boundary markers: Absent; (3) Capability inventory: ssh, curl, pip, sudo; (4) Sanitization: None detected.

vps-proxy-tunnel