The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to download and execute a shell script directly from Lambda Labs' official domain (lambdalabs.com) to install or update the system stack.
[COMMAND_EXECUTION]: Includes various administrative and system-level commands, such as sudo reboot, sudo apt install, and Docker operations, which are intended for managing the environment of the launched GPU instances.
[EXTERNAL_DOWNLOADS]: The documentation and scripts reference several external resources, including Python packages from PyPI, machine learning models from Hugging Face, and setup scripts from vendor-owned domains.
[INDIRECT_PROMPT_INJECTION]: The skill includes automation scripts for executing commands on remote instances and streaming back logs (e.g., via paramiko). This ingestion of external data constitutes a potential attack surface, although it is within the intended scope of managing remote infrastructure.
Ingestion points: Remote command output and log files streamed via SSH using the paramiko library in references/advanced-usage.md.
Boundary markers: None explicitly implemented in the provided Python automation scripts.
Capability inventory: Includes remote shell execution (client.exec_command), file system writes, and package installations (pip install).
Sanitization: The scripts do not include specific sanitization for the content returned from remote logs or command outputs before they are printed to the console.

lambda-labs-gpu-cloud