The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The config.yaml file contains hardcoded local system paths to sensitive SSH private keys (C:/Users/qian/.ssh/id_ed25519). Distributing a configuration with specific user paths and high-privilege access credentials (e.g., username: root) constitutes a credential exposure risk and suggests the accidental inclusion of a private development configuration.
[COMMAND_EXECUTION]: The skill's primary purpose is to facilitate arbitrary command execution on remote servers via SSH. While this is the intended functionality, it provides a powerful primitive that could be abused if the agent is manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from remote servers which could lead to indirect prompt injection.
Ingestion points: The skill reads remote command outputs, log files, and directory listings (e.g., ls -la, cat, training logs) into the agent's context.
Boundary markers: None detected in the provided scripts or instructions to delimit remote content from system instructions.
Capability inventory: Extensive capabilities including SSH command execution, file transfer, and tunnel creation across connect/, deploy/, and tunnel/ modules.
Sanitization: No evidence of sanitization or filtering of remote output before it is processed by the agent.

ssh-dev-suite