clinical-trial-matcher

name: 'clinical-trial-matcher' description: 'Matches patient profiles to open clinical trials using vector similarity and inclusion/exclusion criteria. Use when a user provides patient data and asks for relevant trials.' measurable_outcome: Execute skill workflow successfully with valid output within 15 minutes. allowed-tools: - read_file - run_shell_command

Clinical Trial Matcher

This skill matches a patient's clinical profile against a database of active clinical trials (ClinicalTrials.gov).

When to use this skill

• When the user provides a patient case summary, medical history, or specific diagnosis.
• When the user explicitly asks to "find trials" or "recruit patients".
• When evaluating eligibility criteria for a specific trial.

How to use it

1. Extract Patient Data: Identify key demographics (age, sex), primary diagnosis, stage of disease, and prior lines of therapy from the user's input.
2. Search Vector Database:
   • Generate an embedding for the patient summary.
   • Query the trials_vector_db for top 20 semantic matches.
3. Filter by Logic:
   • Apply hard filters: status=Recruiting, phase!=Terminated.
   • Check inclusion/exclusion criteria (e.g., "Must have EGFR mutation", "No prior immunotherapy").
4. Format Output:
   • Present matches in a structured list: Trial ID (NCT number), Title, Phase, and "Why it fits" (matching logic).
   • Highlight any potential exclusion criteria that need verification.