The Agent Skills Directory

[SAFE]: The skill interacts exclusively with official NVIDIA endpoints, including health.api.nvidia.com for inference and api.ngc.nvidia.com for database downloads, which is appropriate for a vendor-authored skill.
[COMMAND_EXECUTION]: Shell commands provided for local deployment are legitimate, focusing on container orchestration with Docker and parallel database retrieval using standard tools like aria2c.
[CREDENTIALS_SAFE]: The skill implements best practices for secret management, instructing users to provide NGC_API_KEY through environment variables or local .env files, avoiding hardcoded credentials.
[FILE_SYSTEM_SECURITY]: The included scripts/hosted_search.py utility demonstrates high security maturity by enforcing strict POSIX file permissions (0600) and directory permissions (0700), and using temporary staging areas to ensure atomic and private writes.
[DATA_VALIDATION]: The Python client performs strict validation of input protein sequences against the standard amino acid alphabet, effectively mitigating potential injection attacks at the application layer.

msa-search-nim