Core Workflow

Illustrative workflow; example queries, dates, counts, and numerical findings are not executed review results. Run commands from this skill directory, using paths to your review workspace.

All seven phases in full: planning and scoping, systematic search, screening and selection, data extraction and quality assessment, synthesis and analysis, citation verification, and document generation.

Core Workflow

Literature reviews follow a structured, multi-phase workflow:

Phase 1: Planning and Scoping

1. Define Research Question: Use PICO framework (Population, Intervention, Comparison, Outcome) for clinical/biomedical reviews
2. Example: "What is the efficacy of CRISPR-Cas9 (I) for treating sickle cell disease (P) compared to standard care (C)?"

1. Establish Scope and Objectives:
2. Define clear, specific research questions
3. Determine review type (narrative, systematic, scoping, meta-analysis)
4. Set boundaries (time period, geographic scope, study types)

1. Develop Search Strategy:
2. Identify 2-4 main concepts from research question
3. List synonyms, abbreviations, and related terms for each concept
4. Plan Boolean operators (AND, OR, NOT) to combine terms
5. Select complementary bibliographic databases and registries appropriate to the question
6. **Optionally use parallel-web (parallel-cli search) for initial scoping** to quickly gauge the landscape before formal database searches

1. Set Inclusion/Exclusion Criteria:
2. Date range (e.g., the explicitly historical interval 2015-2024)
3. Language (typically English, or specify multilingual)
4. Publication types (peer-reviewed, preprints, reviews)
5. Study designs (RCTs, observational, in vitro, etc.)
6. Document all criteria clearly

Phase 2: Systematic Literature Search

1. Multi-Database Search:

Select bibliographic databases and registries appropriate for the domain. Use optional web search for scoping and supplementary discovery; record it separately from database exports.

Supplementary web discovery (parallel-web):

• Use parallel-cli search with academic domain filtering for broad scholarly coverage
• Consider academic-focused and general searches; ranked samples do not guarantee complete retrieval ```bash mkdir -p sources # Illustrative ranked sample; not a complete search parallel-cli search "your research topic" -q "keyword1" -q "keyword2" \ --json --max-results 10 --excerpt-max-chars-total 27000 \ --include-domains "scholar.google.com,arxiv.org,pubmed.ncbi.nlm.nih.gov,semanticscholar.org,biorxiv.org,medrxiv.org,ncbi.nlm.nih.gov,nature.com,science.org,ieee.org,acm.org,springer.com,wiley.com,cell.com,pnas.org,nih.gov" \ -o sources/litreview_<topic>-academic.json

# General search for supplementary sources parallel-cli search "your research topic" -q "keyword1" -q "keyword2" \ --json --max-results 10 --excerpt-max-chars-total 27000 \ -o sources/litreview_<topic>-general.json ```

• Use parallel-cli extract --full-content to request content from accessible paper URLs; check for truncation and retrieval errors ``bash parallel-cli extract "https://arxiv.org/abs/1706.03762" --full-content --json

Biomedical & Life Sciences:

• Search PubMed via its interface or E-utilities; PMC has different full-text coverage
• Search bioRxiv/medRxiv interfaces, or retrieve date-range API metadata and explicitly filter it
• See database_strategies.md for exact requests, authentication, and export limits

General Scientific Literature:

• Search arXiv via direct API (preprints in physics, math, CS, q-bio)
• Search Semantic Scholar using its documented relevance or bulk API as appropriate
• Use Google Scholar manually for supplementary discovery and citation chaining; preserve search and screening limits

Biological entity databases can inform background, but do not count as bibliographic sources simply because they contain references.

1. Document Search Parameters: ```markdown ## Search Strategy

### Database: PubMed

• Date searched: 2024-10-25
• Date range: 2015-01-01 to 2024-10-25
• Search string: `` ("CRISPR"[Title] OR "Cas9"[Title]) AND ("Anemia, Sickle Cell"[MeSH Terms] OR "SCD"[Title/Abstract]) AND 2015:2024[Publication Date]
• Results: 247 articles ```

Repeat for each database searched.

1. Export and Aggregate Results:
2. Export results in JSON format from each database
3. Normalize raw provider responses into a JSON array, preserving raw exports and provenance separately
4. Use scripts/search_databases.py for post-processing: ``bash python scripts/search_databases.py combined_results.json \ --deduplicate \ --format markdown \ --output aggregated_results.md

Phase 3: Screening and Selection

1. Deduplication: ``bash python scripts/search_databases.py results.json --deduplicate --format json --output unique_results.json
2. Keeps first normalized DOI; title fallback compares DOI-less records only. Review matches and retain raw duplicates for enrichment
3. Document number of duplicates removed

1. Title Screening:
2. Review all titles against inclusion/exclusion criteria
3. Exclude obviously irrelevant studies
4. Document number excluded at this stage

1. Abstract Screening:
2. Read abstracts of remaining studies
3. Apply inclusion/exclusion criteria rigorously
4. Document reasons for exclusion

1. Full-Text Screening:
2. Obtain full texts of remaining studies
3. Conduct detailed review against all criteria
4. Document specific reasons for exclusion
5. Record final number of included studies

1. Create PRISMA Flow Diagram:
2. Use the official template matching new/updated review and source types.
3. Track records identified and removed before screening, records screened and excluded, reports sought/not retrieved/assessed, excluded reports with reasons, and included studies plus reports of those studies.
4. Reconcile counts at each transition; do not invent one included-study count from the number of publications. Maintain a study-to-report map.

Phase 4: Data Extraction and Quality Assessment

1. Extract Key Data from each included study:
2. Study metadata (authors, year, journal, DOI)
3. Study design and methods
4. Sample size and population characteristics
5. Key findings and results
6. Limitations noted by authors
7. Funding sources and conflicts of interest

1. Assess Risk of Bias and Certainty:
2. Use a tool appropriate to the design and question: RoB 2 for randomized trial results; ROBINS-I for nonrandomized intervention results; AMSTAR 2 for systematic reviews of healthcare interventions. Other designs require their own appraisal tool.
3. Preserve domain judgments and rationale. RoB 2 uses low risk, some concerns, or high risk; do not force every tool onto a common numerical scale.
4. GRADE assesses a body of evidence for each outcome (high/moderate/low/very low), not a generic grade for each individual study. See the Cochrane certainty guidance.
5. Pre-specify how bias affects synthesis; do not invent post-hoc exclusions.

1. Organize by Themes:
2. Identify 3-5 major themes across studies
3. Group studies by theme (studies may appear in multiple themes)
4. Note patterns, consensus, and controversies

Phase 5: Synthesis and Analysis

1. Create Review Document from template: ``bash cp assets/review_template.md my_literature_review.md

1. Write Thematic Synthesis (NOT study-by-study summaries):
2. Organize Results section by themes or research questions
3. Synthesize findings across multiple studies within each theme
4. Compare and contrast different approaches and results
5. Identify consensus areas and points of controversy
6. Highlight the strongest evidence

Fictitious structure example (replace every count and percentage with extracted evidence): ```markdown #### 3.3.1 Theme: CRISPR Delivery Methods

Multiple delivery approaches have been investigated for therapeutic gene editing. Viral vectors (AAV) were used in 15 studies^1-15^ and showed high transduction efficiency (65-85%) but raised immunogenicity concerns^3,7,12^. In contrast, lipid nanoparticles demonstrated lower efficiency (40-60%) but improved safety profiles^16-23^. ```

1. Critical Analysis:
2. Evaluate methodological strengths and limitations across studies
3. Assess quality and consistency of evidence
4. Identify knowledge gaps and methodological gaps
5. Note areas requiring future research

1. Write Discussion:
2. Interpret findings in broader context
3. Discuss clinical, practical, or research implications
4. Acknowledge limitations of the review itself
5. Compare with previous reviews if applicable
6. Propose specific future research directions

Phase 6: Citation Verification

CRITICAL: All citations must be verified for accuracy before final submission.

1. Verify All DOIs: ``bash python scripts/verify_citations.py my_literature_review.md

This script:

• Extracts all DOIs from the document
• Checks DOI registration via the Handle API
• Retrieves available Crossref metadata and preserves unresolved errors
• Generates verification report
• Does not verify claim support, publisher accessibility, or final citation style

1. Review Verification Report:
2. Check for any failed DOIs
3. Verify author names, titles, and publication details match
4. Correct any errors in the original document
5. Resolve errors against the source; do not delete a reference solely because a service timed out or lacks its metadata

1. Format Citations Consistently:
2. Choose one citation style and use throughout (see references/citation_styles.md)
3. Common styles: APA, Nature, Vancouver, Chicago, IEEE
4. Use complete metadata with a CSL processor; the script only provides metadata previews
5. Ensure in-text citations match reference list format

Phase 7: Document Generation

1. Generate PDF: ``bash python scripts/generate_pdf.py my_literature_review.md \ --bibliography my_literature_review.bib --csl styles/apa.csl \ --output my_review.pdf

Supply an existing .bib and .csl for that example; they are not bundled. [@key] citations are formatted, while manually typed references remain unchanged.

Options:

• --csl / --citation-style: existing local CSL path/name; omitted uses Pandoc default
• --bibliography: bibliography file; sibling .bib also auto-detected
• --output: PDF destination
• --no-toc: Disable table of contents
• --no-numbers: Disable section numbering
• --check-deps: Check if pandoc/xelatex are installed

1. Review Final Output:
2. Check PDF formatting and layout
3. Verify all sections are present
4. Ensure citations render correctly
5. Check that figures/tables appear properly
6. Verify table of contents is accurate

1. Quality Checklist:
2. [ ] DOI registration checked; metadata identity and claim support reviewed manually
3. [ ] Citations formatted consistently
4. [ ] PRISMA flow diagram included (for systematic reviews)
5. [ ] Search methodology fully documented
6. [ ] Inclusion/exclusion criteria clearly stated
7. [ ] Results organized thematically (not study-by-study)
8. [ ] Quality assessment completed
9. [ ] Limitations acknowledged
10. [ ] References complete and accurate
11. [ ] PDF generates without errors