The Agent Skills Directory

[SAFE]: No security issues or malicious behaviors were identified in the skill instructions or accompanying Python scripts.
[PROMPT_INJECTION]: The instructions focus entirely on scientific modeling workflows. There are no attempts to override system prompts, bypass safety filters, or manipulate agent behavior.
[DATA_EXFILTRATION]: The skill does not access sensitive local paths (such as SSH keys or credentials) and does not perform network operations. Data processing is confined to user-provided input files.
[REMOTE_CODE_EXECUTION]: The skill relies on well-known scientific libraries (numpy, scipy). It does not download or execute remote code from untrusted sources.
[OBFUSCATION]: The Python scripts and documentation are clear, well-commented, and free of hidden payloads, Base64 encoding, or homoglyph-based deception.
[COMMAND_EXECUTION]: While the skill utilizes shell access to run its local Python scripts, it uses strict input parsing (e.g., float validation) which prevents command injection via malformed data files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data (CSV/TSV files). However, the scripts apply strict numerical parsing and validation, effectively sanitizing inputs before they are used in calculations or reported back to the agent.

pkpd-modeling