answering-research-questions_skill

Overview

This skill orchestrates a systematic literature research workflow from query parsing through synthesis and final consolidation. It focuses on reproducible, trackable steps: Search → Evaluate → Traverse → Synthesize. The goal is to produce a clear SUMMARY.md, a machine-friendly relevant-papers.json, and a complete papers-reviewed.json audit trail.

How this skill works

It parses the user's research question into keywords, data types, and constraints, then creates a dated research folder and core tracking files. It runs a PubMed-based search, scores and extracts data from each paper, follows citations forward and backward for highly relevant items, and consolidates findings into structured markdown and JSON outputs. Checkpoints and periodic reports keep the user informed and allow early stopping or continuation.

When to use it

• When you need a reproducible literature review with provenance and statistics
• When extracting quantitative data (IC50, KD, EC50) or methodological details across many papers
• When you want automated citation traversal to find related high-value papers
• For coordinated programmatic access to results via relevant-papers.json
• When you need clear checkpointing during long searches

Best practices

• Start by clarifying data types, time range, species, and any target compounds
• Provide an email for Unpaywall API requests to locate open-access full text
• Use the rubric threshold (score ≥7) to limit deep review effort
• Create auxiliary files (evaluated-papers.json, TOP_PRIORITY_PAPERS.md) for searches >100 papers
• Run checkpoints every 50 papers or every 5 minutes to avoid runaway processing

Example use cases

• Compare IC50 and selectivity profiles for a class of kinase inhibitors
• Collect synthesis methods and yields across a set of lead compounds
• Build a prioritized list of papers for experimental replication or meta-analysis
• Discover overlooked relevant literature via forward/backward citation traversal
• Produce a reproducible research bundle (SUMMARY.md + relevant-papers.json) for collaborators or reviewers