The Agent Skills Directory

[SAFE]: No malicious behavior or security risks were identified. The skill acts as an instructional guide for academic research planning.- [PROMPT_INJECTION]: The content consists of natural instructional language and templates for experiment design. It does not contain instructions to override system prompts, bypass safety filters, or ignore previous instructions.- [DATA_EXFILTRATION]: There is no evidence of sensitive data access, credential harvesting, or unauthorized network communication. The skill mentions computational budgets in an abstract, planning context only.- [EXTERNAL_DOWNLOADS]: Several standard Python packages (e.g., PyTorch, Transformers, NumPy) are mentioned within a reproducibility checklist for users to document their environment. These are for reference and documentation purposes and do not trigger any automated installations or remote script executions.- [REMOTE_CODE_EXECUTION]: No remote code patterns, shell command execution, or dynamic evaluation of external content were found.

experiment-design