clinical-trial-landscape

Files (4)

• SKILL.md
• openai.yaml · agents/
• handler.py · scripts/
• skill.json

SKILL.md (first lines)

# Clinical Trial Landscape

Translate non-English disease and intervention concepts into precise search terms while preserving drug names, gene symbols, and requested status scope.

Use `clinical_trials_preview` for a focused answer using no more than ten records. Cite NCT IDs and returned ClinicalTrials.gov links for material registry claims. Keep recruitment status, study design, registered endpoints, posted results, and linked publications distinct.

Use `clinical_trials_collect` only through a researcher-approved workflow when the user needs a durable landscape. Preserve the exact condition, intervention, status and study filters, source order, NCT IDs, retrieval time, manifest, and output paths.

Use `clinical_trial_results_preflight` when the user gives an exact NCT ID and asks about posted results, outcome values, statistical analyses, participant flow, baseline, or adverse events. It verifies that tabular results exist and exposes only study identity and bounded counts before approval.

Use `clinical_trial_results_review` only through the researcher-approved exact-results workflow. Preserve source order, group mappings, units, denominators, time frames, submitted p-values and confidence intervals, adverse-event group totals, protocol and SAP documents, linked publications, retrieval time, manifest, and output paths. Do not generate a custom treatment-effect, safety, certainty, or study-quality score.

A registry entry is not evidence that an intervention is effective or safe. Registered outcomes are not measured results, overall status may differ from individual site status, and missing results must not be interpreted as study failure. Posted tables are sponsor- or investigator-submitted summaries, not independent validation. Review the live record, group definitions, analysis population, protocol, analysis plan, publications, and relevant regulatory sources before drawing translational conclusions.

From phenixace/Molemo_WorkBench, MIT.

More from phenixace/Molemo_WorkBench