The Agent Skills Directory

[SAFE]: The skill operates entirely within a localized 'research/' directory structure, managing research proposals, experiment plans, and tasks. It does not attempt to access sensitive system files or environment variables.
[SAFE]: No network operations or data exfiltration patterns were identified. The skill does not use commands like curl or wget to communicate with external servers.
[SAFE]: The coding style and workflow documentation encourage secure and reproducible practices, such as using loggers instead of print statements and centralizing configuration management.
[SAFE]: Although the skill generates task lists that may include shell commands for experiments (e.g., training scripts), these are presented as templates for legitimate research activities and do not constitute unauthorized command execution.

research-experiment