The Agent Skills Directory

[SAFE]: The skill implements strong isolation by using ephemeral sandboxes and dedicated state directories for its operations, preventing interference with the host environment.
[SAFE]: It enforces a strict security policy that automatically identifies and excludes sensitive data, such as local secrets and credentials, from being read or committed during the optimization process.
[SAFE]: Access controls are well-defined, restricting the 'Optimizer' role to specific directory writes while keeping the 'Evaluator' role strictly read-only to ensure integrity.
[NO_CODE]: The skill provides procedural guidance and documentation rather than executable scripts, minimizing the risk of direct malicious code execution.

automake