getting-started_skill

Overview

This skill introduces systematic literature search and review workflows for finding, screening, extracting, and following citations in published papers. It focuses on reproducible, test-driven methods to locate papers with the exact data you need rather than broad topical discovery. The skill coordinates searches, screening rubrics, data extraction, and organized research sessions.

How this skill works

The workflow orchestrates search → screen → extract → traverse: run targeted searches (PubMed, Semantic Scholar), build and validate screening rubrics, perform two-stage screening (abstract then full text), extract methods/results, and follow citations selectively. It manages research session folders with summaries, deduplication tracking by DOI, cached abstracts, and checkpointed progress updates during large reviews.

When to use it

• Conducting a systematic literature review focused on extracting specific measurements, methods, or results
• Running large-scale screening of 50+ papers using parallel subagents
• Building and validating screening criteria before bulk screening
• Tracing backward/forward citations to expand a focused evidence base
• Finding open-access copies of paywalled papers for extraction and archiving

Best practices

• Prioritize precision: craft queries to find papers likely to contain the exact data you need
• Design and test screening rubrics on a small set before bulk processing
• Deduplicate by DOI at the start and log all reviewed papers in papers-reviewed.json
• Cache abstracts and saved metadata to allow re-screening when rubrics change
• Checkpoint progress regularly (report every 10 papers, ask to continue every 50)

Example use cases

• Extracting numeric outcome measures and methods from clinical trial reports for a meta-analysis
• Screening thousands of abstracts with parallel subagents, then deep-diving on selected papers
• Building a validated rubric to identify studies that report a specific assay or measurement
• Following key citations from a seminal paper to build a focused citation network
• Finding free PDFs of paywalled articles via Unpaywall before extracting supplementary data