The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the tfy CLI tool and a helper script tfy-api.sh to manage deployments. These are standard operations for the skill's purpose of interfacing with the TrueFoundry platform.
[EXTERNAL_DOWNLOADS]: The skill installs the truefoundry Python package from standard registries and provides instructions for downloading model artifacts from Hugging Face. These are legitimate actions for an ML deployment tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied model IDs and metadata from the Hugging Face hub. It includes proactive security measures, such as explicit warnings regarding the --trust-remote-code flag and instructions to avoid fetching external release pages at runtime to prevent instruction injection.
[CREDENTIALS_SAFE]: The skill uses environment variables and .env files for managing API keys, which is documented as a safe practice for secret management in this context. The helper script tfy-api.sh includes validation to prevent path traversal when making API calls.

truefoundry-llm-deploy