The Agent Skills Directory

[SAFE]: The skill follows security best practices by delegating authentication to the Membrane platform via membrane login and membrane connect, ensuring that the agent does not directly handle or store sensitive user credentials.
[SAFE]: All external dependencies, such as @membranehq/cli, originate from the vendor's official package registry and are necessary for the skill's primary function.
[SAFE]: The dynamic creation of actions (membrane action create) is a core feature of the Membrane platform's orchestration layer and is used here for its intended purpose of automating W&B workflows.
[SAFE]: Network operations are restricted to the vendor's domain (getmembrane.com) and the official Weights & Biases service (wandb.ai).

wb