Literature Synthesis skills for AI agents

Activate this skill when the user wants to see the shape of a field rather than pool its results: who cites whom, where the evidence clusters, and where it is absent. Triggers on "citation network," "co-citation analysis," "bibliographic coupling," "evidence gap map," "evidence map," "seminal papers," "bibliometric analysis," "VOSviewer," "citation chasing," "mapping the literature," or "literature review visualization." Covers building citation graphs from reference lists (including offline from PDFs), computing centrality and clusters, constructing intervention-by-outcome evidence gap maps, identifying landmark and outlier works, and presenting the map without overclaiming.

Activate this skill when the user needs to judge the trustworthiness of included studies and rate certainty in a body of evidence. Triggers on "risk of bias," "RoB 2," "ROBINS-I," "Newcastle-Ottawa," "critical appraisal," "CASP checklist," "GRADE," "certainty of evidence," "summary of findings table," "CERQual," "quality assessment," or "literature review appraisal." Covers matching the appraisal tool to the study design, applying signalling questions consistently, rating GRADE certainty across the downgrading and upgrading domains, appraising qualitative studies, assessing confidence with GRADE-CERQual, and recording judgments so a reader can retrace them.

Activate this skill when the user wants to pool quantitative results across studies or needs to judge whether pooling is defensible. Triggers on "meta-analysis," "effect size," "standardized mean difference," "pooled odds ratio," "fixed effect vs random effects," "heterogeneity," "I squared," "forest plot," "funnel plot," "publication bias," "Egger's test," "metafor," or "literature review statistics." Covers choosing and computing effect sizes, inverse-variance pooling under fixed-effect and random-effects models, quantifying heterogeneity, reading forest and funnel plots, small-study effects, and the conditions under which a meta-analysis should not be done.

Activate this skill when the user hands you a corpus they already have (PDFs, extracted text, reading notes, reference-manager exports) and wants it synthesized with no web access and no database searching. Triggers on "offline literature synthesis," "literature synthesis," "literature review," "synthesize these papers," "what do these PDFs say," "cross-paper matrix," "evidence table from my folder," or "summarize my reading notes." Covers corpus inventory, reading order, structured extraction, cross-study matrices, and writing a synthesis in which every claim traces back to a document and page in the corpus.

Activate this skill when the user wants to identify what a body of literature has not answered and turn that into fundable, answerable research questions. Triggers on "research gap," "gap analysis," "gaps in the literature," "future research," "research agenda," "research questions from a literature review," "proposal from a literature synthesis," "under-researched," or "what is missing in the literature." Covers a typology of gaps (population, method, theory, context, evidence, measurement), writing gap statements backed by counts and citations, distinguishing an absence of studies from an absence of good studies, and converting gaps into prioritized questions and proposal aims.

Activate this skill when the user is moving from a completed search to an included-study set and needs to screen records, resolve disagreements, and extract data into evidence tables. Triggers on "title and abstract screening," "full-text screening," "dual screening," "inter-rater agreement," "Cohen's kappa," "data extraction form," "evidence table," "reference management," "deduplication," "Rayyan," "Covidence," or "literature review screening." Covers two-stage screening, calibration and conflict resolution, deduplication and reference tracking, piloted extraction forms, and building characteristics-of-included-studies and results tables.

Activate this skill when the user is planning a systematic review, scoping review, or other structured literature review and needs a protocol before any searching begins. Triggers on "systematic review protocol," "PICO question," "PEO question," "inclusion and exclusion criteria," "search strategy," "PRISMA flow diagram," "PROSPERO registration," "literature review plan," "review team roles," or "literature synthesis protocol." Covers framing an answerable question, writing eligibility criteria screeners can apply consistently, documenting reproducible database searches, planning the PRISMA flow, registering the protocol, and assigning roles and timelines to a review team.

Activate this skill when the user has qualitative, mixed-methods, or heterogeneous quantitative studies that cannot be pooled statistically and needs a rigorous non-statistical synthesis. Triggers on "thematic synthesis," "narrative synthesis," "framework synthesis," "meta-ethnography," "qualitative evidence synthesis," "coding studies," "theme development," "synthesis without meta-analysis," "mixed-methods review," or "literature review themes." Covers line-by-line coding, descriptive and analytical theme development, best-fit framework synthesis, the structure of a narrative synthesis, and handling heterogeneity across qualitative and mixed evidence.

Activate this skill when the user has finished extraction and appraisal and needs to turn tables into a review manuscript, thesis chapter, or report. Triggers on "writing a literature review," "review structure," "results section of a systematic review," "hedged claims," "conflicting studies," "structured abstract," "plain language summary," "PRISMA checklist," "literature synthesis writing," or "how to write up findings." Covers the standard architecture of a review, moving from evidence tables to prose, calibrating claims to certainty, reporting disagreement among studies without averaging it away, and writing abstracts and plain-language summaries that survive being quoted.