The Agent Skills Directory

[COMMAND_EXECUTION]: The skill recommends launching agents on remote servers in autonomous modes (e.g., codex --yolo), which can lead to unintended or high-risk system changes without human oversight.
[COMMAND_EXECUTION]: Provides instructions for managing servers via root SSH access, granting full administrative control over the host systems.
[DATA_EXPOSURE]: Directs the agent to treat ~/infrastructure.md as a primary data source, which contains sensitive information such as server IP addresses and infrastructure metadata.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external, potentially untrusted data.
Ingestion points: ~/infrastructure.md, journalctl system logs, and ~/.hermes/config.yaml.
Boundary markers: None provided; the agent is not instructed to distinguish between configuration data and embedded instructions.
Capability inventory: The skill allows for arbitrary shell command execution on remote hosts, file system access, and service management via SSH (referenced in SKILL.md).
Sanitization: No evidence of input validation or sanitization when processing logs or configuration files.

vps-server-management