clinical-trial-matching-skill

Use when contextualizing a clinical trial's expected efficacy and comparing against current standard-of-care for the patient's cancer type and treatment line. Triggers when the parent clinical-trial-matching skill needs efficacy snapshots and vs-SoC head-to-head data per decision path. Replaces v1.7.x efficacy_lookup.py + efficacy_database.json + soc_benchmarks.json (which had only 1 CRC SoC entry, leaving most CRC patients with "vs SoC: not available" in their report).

Use when matching an oncology patient (esp. Chinese patients) to clinical trials and generating a decision-grade HTML report. Triggers on phrases like "临床试验匹配", "trial matching", "shortlist trials for this patient", "MTB trial section". Coordinates dual-source retrieval (ClinicalTrials.gov + ChiCTR) and 4 LLM subskills for gating, risk, efficacy, and decision synthesis.

Use when synthesizing per-trial gating + risk + efficacy outputs into a Top-N decision report with diversity bucketing, Goals-of-Care trigger, and vs-SoC head-to-head. Triggers as the final synthesis step in clinical-trial-matching after trial-gater + trial-risk-annotator + trial-efficacy-contextualizer have run. Replaces v1.7.x synthesis/decision_paths.py + goals_of_care.py + consistency_check.py.

Use when evaluating one or more clinical trials' eligibility against a specific patient profile, criterion-by-criterion. Triggers when the parent clinical-trial-matching skill needs gating verdicts (match / conditional / exclude) with R1-R5 hard rules applied. Replaces v1.7.x scoring/gating.py + regex-based rule engine.

Use when annotating a clinical trial's mechanism-specific risks for a specific patient and cancer type. Triggers when the parent clinical-trial-matching skill needs risk narratives for the report. Replaces v1.7.x risk_lookup.py + risk_taxonomy.json (which leaked PDAC-specific risk text onto CRC reports because the taxonomy was keyed on mechanism only, not (mechanism × cancer)).