spb/autoscholar

1---23mode: review4discipline: medicine5title: "<Working title: a literature review>"6---78# Research brief — <working title> (review)910## 1. Research objective and scope1112- **Review question.** <PICO-style or conceptual question.>13- **Type.** <systematic | scoping | narrative/critical>.14- **Scope.** <years, languages, study designs, populations, outcomes>.1516## 2. Domain, mode and methodology1718- Discipline: <…>. Primary mode: review. Reporting standard: <PRISMA 2020 / PRISMA-ScR / narrative with explicit protocol>.1920## 3. Contribution and novelty2122- <Why a new synthesis is needed; existing reviews to compare with (cite).>2324## 4. Search protocol and source rules2526- Databases: <PubMed, OpenAlex, Crossref, Semantic Scholar, arXiv>. Search strings: <exact strings per database>. Date of search recorded.27- Inclusion/exclusion criteria: <…>. Screening: title/abstract then full text (open-access when available); record every decision with a reason in `data/screening.csv`.28- Every included study added through the verification tool (DOI/PMID); target ≥ 25 verified sources.2930## 5. Evidence extraction3132- Data items: <design, sample, intervention/exposure, outcomes, effect sizes, risk-of-bias items>. Store in `results/extraction.csv`.33- Quality/risk-of-bias assessment: <tool>.3435## 6. Procedure36371. Run searches; export hit counts per database (`data/manifest.json` describes the screening records).382. Screen; extract; assess quality.393. Synthesize: <narrative synthesis / vote counting / meta-analysis if ≥ k comparable studies — with heterogeneity and small-study checks>.404. Produce the flow diagram (counts), evidence tables and summary figures.4142## 7. Robustness and falsification4344- Re-run searches with alternative terms; re-screen a random 10% sample for consistency; sensitivity analyses excluding low-quality studies.4546## 8. Reproducibility and artifacts4748- Search strings, dates, counts, screening and extraction files, synthesis scripts (`code/`), `results/synthesis.md`.4950## 9. Manuscript5152- AutoScholar class; methods section reproducing the protocol; evidence tables; limitations of the evidence base; 10–25 pages.5354## 10. Review and revision5556- Reviewer focus: search completeness, selection bias, extraction errors, overgeneralized conclusions.5758## 11. Limitations, ethics, legal5960- Metadata and abstracts only where full texts are paywalled; no republication of copyrighted text beyond short quotes. Medical conclusions are not clinical advice.6162## 12. Acceptance criteria6364- Protocol executed as written; counts reported; ≥ 25 verified sources; extraction table complete; synthesis consistent with the extracted data.65