The Agent Skills Directory

[SAFE]: The skill is authored by a reputable academic entity (mims-harvard) and focuses on systematic toxicological research using legitimate data sources.
[COMMAND_EXECUTION]: The skill uses Python scripts for data processing and statistical analysis. It relies on standard, industry-standard libraries including pandas, scipy, statsmodels, and matplotlib to handle biological data.
[EXTERNAL_DOWNLOADS]: The skill references and retrieves information from official government and research domains, such as fda.gov, aopwiki.org, ctdbase.org, and nlm.nih.gov. These are recognized institutional sources for medical and chemical data.
[DYNAMIC_EXECUTION]: The instructions include logic to generate and run Python code via bash to perform enrichment analysis and signal detection. This is consistent with the skill's intended purpose of performing complex calculations on retrieved datasets.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The agent processes data returned from external tool calls to scientific databases (AOPWiki, FAERS, DailyMed, CTD) mentioned in SKILL.md.
Boundary markers: Not explicitly defined in the workflow instructions.
Capability inventory: The agent is authorized to execute bash commands to run locally generated Python scripts in SKILL.md.
Sanitization: No specific filtering for tool output content is specified; however, the data sources are trusted academic and regulatory repositories.

tooluniverse-toxicology