Medical Literature Report

What problem it addresses

Medical literature briefings often fail at three points: paper selection is inaccurate, source data are not verified, and slides mix the authors' conclusions with the presenter's inferences. This skill is for Chinese medical reporting across specialties. Inputs can include disease, population, intervention or exposure, test, biomarker, outcome, DOI/PMID, journal tier, publication window, article type, and deliverables. It also separates hard criteria from preferred criteria so ranking preferences do not become inclusion filters.

How it works

The skill reads screening, acquisition, translation, study-design appraisal, reporting, and evidence-interpretation references, then uses article_inventory.py to inventory PDFs, Office files, and deliverables, and package_deliverables.py to create a verified delivery package. The workflow first identifies article type and study design, then searches primary sources and checks journal status, metrics, guidelines, and access. It builds a candidate matrix instead of judging titles alone, verifies bibliographic identity, sample size, methods, outcomes, full-text status, and figure completeness, and obtains lawful full text and supplements. It creates a source fact sheet with numbers, units, effect estimates, confidence intervals, P values, time points, eligibility criteria, outcome definitions, and limitations, then performs consistent Chinese translation or close reading. It applies design-specific appraisal, distinguishes risk of bias, reporting quality, external validity, and practical relevance, and builds a PPT claim spine from background, gap, objective, design, methods, results, interpretation, limitations, and implications. Deliverables may include a candidate table, provenance log, translation document, appraisal report, Chinese PPTX, speaker notes, discrepancy log, and hash-verified archive.

Boundaries

It is suitable for research, education, departmental reporting, and cross-team interpretation, not patient-specific medical advice. The skill avoids fabricated data, unsupported causal claims, and presenting exploratory cutoffs as validated clinical decision limits. Presenter inference, external evidence, and source conclusions are labeled separately. When the user does not define constraints, the skill infers conservatively and states material assumptions. Final outputs should not contain patient-identifiable or confidential data.