The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or security risks were detected. The skill serves as legitimate documentation for the SkyPilot orchestration tool.
[EXTERNAL_DOWNLOADS]: The skill references standard installation of packages from official registries and provides links to official documentation and repositories.
Fetches the 'skypilot' package and ML libraries like 'torch', 'transformers', and 'deepspeed' via pip.
References official GitHub repositories and documentation sites (skypilot.co, github.com/skypilot-org).
[COMMAND_EXECUTION]: Provides extensive examples of shell commands for using the SkyPilot CLI to manage cloud clusters and execute training tasks.
Commands like sky launch, sky jobs launch, and sky serve up are documented for their intended infrastructure management purposes.
Includes examples of using torchrun and deepspeed for distributed training orchestration.
[CREDENTIALS_UNSAFE]: The documentation correctly identifies how to manage sensitive credentials using environment variables (e.g., HF_TOKEN, WANDB_API_KEY) and platform-specific secret management features, promoting secure practices over hardcoding.
[DATA_EXFILTRATION]: No suspicious network activity or data exfiltration patterns were found. Network operations are limited to standard cloud provider interactions and documentation references.

skypilot-multi-cloud-orchestration