The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the tooluniverse library from a public package registry as specified in README.md and SKILL.md to function correctly.
[DATA_EXFILTRATION]: The provided scripts (python_implementation.py, trial_pipeline.py) use the tooluniverse library to perform network operations, querying multiple external APIs and databases including PubMed, ClinicalTrials.gov, DrugBank, and Open Targets to gather clinical and regulatory data.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests large amounts of untrusted data from external research databases and interpolates this content into reasoning tasks and feasibility reports without explicit boundary markers or sanitization.
Ingestion points: Untrusted data enters the agent context through various search tools like PubMed_search_articles, OpenTargets_get_disease_id_description_by_name, and search_clinical_trials as documented in SKILL.md and implemented in the Python scripts.
Boundary markers: The instructions lack the use of delimiters or specific guidance for the agent to ignore or isolate instructions that may be embedded within the retrieved research results.
Capability inventory: The skill includes tools for fetching external data and scripts capable of writing markdown files to the local file system (python_implementation.py).
Sanitization: There is no evidence of automated sanitization, filtering, or validation of the data retrieved from external sources before it is processed by the AI agent.

tooluniverse-clinical-trial-design