literature-and-source-use
Guides finding, mapping, synthesizing, and citing the literature with integrity — critical synthesis, fair engagement with rival work, paraphrase and quotation discipline, and rights and plagiarism ch
SKILL.mdsubagents/research-writing-advisor/skills/literature-and-source-use/SKILL.md
GitHubLiterature And Source Use
Purpose
This skill guides finding, mapping, synthesizing, and citing the literature with integrity. It checks that candidate literature is mapped and searched iteratively and organized as a critical synthesis, that a rival is read for its strongest meaning before disagreement, that ideas are paraphrased and restructured rather than copied, that quotation is reserved for when the wording is the object of attention, and that rights and plagiarism standards are respected so the evidence base is auditable.
When to use
- A field must be mapped and searched, and the literature organized into a topic-based critical synthesis.
- Prior work must be engaged, criticized, or reconciled fairly, reading a rival's strongest meaning first.
- Sources must be paraphrased and cited rather than copied, with quotation used only when wording is the object.
- Legal rights and academic-ethics standards (copyright, plagiarism) apply to reused material.
Procedure
- Map candidate literature by paper type, venue, citation relationships, authors, institutions, dates, and meaningful topic groups, while treating citation counts and search rank as prominence-biased signals rather than quality (P007).
- Introduce complex evidence with the reason or interpretation readers should draw, and provide complete accurate citations and source lists so they can audit evidence and assess the research base (P012).
- Organize literature by a chronology, theme, or method that reveals progress, achievements, limits, gap, and response; within each unit synthesize studies and choose author- or finding-prominent citations for the intended focus (P014).
- Do not closely imitate source prose, figures, or paper design; a citation does not authorize copied wording, so quote a precisely marked permitted span or produce an independently written attributed summary (P016).
- Paraphrase by first understanding and citing the attributable idea, then restructuring its grammar and order while preserving technical terms, qualifiers, meaningful sequence, and original epistemic status (P026).
- Build the literature review continuously as a topic-organized critical synthesis of contributions, limits, relationships, and open questions, preserving displaced early material so reading is not lost (P029).
- Search literature iteratively through complementary paths, using discoveries to refine terminology and criteria, separating broad collection from critical evaluation, and repeating the search as the question evolves (P074).
- Use libraries, librarians, expert reference works, literature reviews, and abstracts to map a field, refine search parameters, find access routes, and select material for close reading (P078).
- Criticize earlier work only when relevant, explain why its original interpretation was reasonable, ground the revision in stronger evidence or known limits, and seek conditional reconciliation before declaring contradiction (P086).
- Use a structured comparison matrix for directly relevant studies to expose differences in problems, methods, results, strengths, limitations, bias, implications, and the new study's extension of the literature (P122).
- Quote only when exact wording is the object of attention, mark and cite it correctly, and otherwise demonstrate understanding through paraphrase rather than filling the paper with quotations (P138).
- Assess both legal rights and academic ethics when owning or reusing intellectual creations because copyright, patents, publication agreements, and plagiarism standards protect different interests (P140).
- Read an important rival first for its strongest coherent meaning and again for critical response; summarize it accurately, inspect its support, and verify every consequential representation before disagreeing (P156).
Inputs
- The literature and sources under discussion, the citation and rights constraints, and the synthesis the writing must build.
- The reasoning offered for the decision under review: the goal, the audience and venue, the draft or plan in place, and any claim of clarity, rigor, or readiness made.
Output
Per finding: name the gap and the principle it engages, give the correction, state the residual trade-off or the referral to make, and end with a concrete next step. Order findings highest-impact first. This skill advises on research writing and presentation; it does not write the paper, section, slides, or talk for the caller, guarantee acceptance or publication, or rule on the domain-science correctness of the research.
Anti-patterns to flag
- A highly-cited paper or top search-rank result is treated as automatically authoritative, or the source list is an unstructured pile with no discernible grouping by venue, author, institution, date, or topic (P007).
- Complex evidence (a statistic, quote, or data point) is dropped into the text with no framing sentence telling the reader what conclusion to draw, or a citation lacks enough detail for the source to be traced (P012).
- The review is a source-by-source list ("Smith found X; Jones found Y...") with no chronological, thematic, or methodological arc, and no stated progress, gap, or response (P014).
- A "paraphrase" keeps the source's sentence structure, figure layout, or worked-example sequence with only words swapped, defended on the grounds that it is cited (P016).
- A paraphrase drops or inflates the original's hedging or qualifiers (e.g., "may reduce" becomes "reduces"), or keeps the source's original grammar and sequence with only synonym substitution (P026).
- The literature review is assembled in a single late pass rather than grown across drafts, and earlier findings are silently dropped or left dangling when new material is inserted (P029).
- The literature search is a single pass frozen at the project's start, never rerun as the question sharpens, with collected material and critically evaluated material left indistinguishable (P074).
- The search relies on one search engine or a convenience sample of papers, with no reference works, review articles, or librarian consultation used to check the search is complete (P078).
- Earlier work is dismissed or attacked without acknowledging why its original interpretation was reasonable at the time, or a flat contradiction is declared where a conditional reconciliation was available (P086).
- Directly relevant studies are compared only in prose, never through a structured matrix, so differences in method, bias, or limitations are never surfaced and the new study's extension of the literature stays unstated (P122).
- The draft is padded with block quotations standing in for the writer's own understanding, or a quotation is used for its content rather than because the exact wording is the point (P138).
- Reused material is cleared on only one axis (e.g., "it's cited, so it's fine") without separately checking copyright, patent, or publication-agreement permission (P140).
- A rival's position is characterized from a weakened or partial reading, or disagreement is voiced against a representation of the rival's argument that was never checked against what the rival actually wrote (P156).
References
Consult ../../references/research-writing-principles-index.md only when a finding's principle needs its full source-grounded statement, or when the issue may belong to a sibling skill and you need to confirm which skill owns it. Consult ../../references/research-writing-evidence-notes.md only when the caller disputes a finding's grounding and you need its source basis.
Provenance
Derived from P007, P012, P014, P016, P026, P029, P074, P078, P086, P122, P138, P140, P156, grounded in the nine distillation-only sources (The Craft of Research; Writing for Computer Science; Writing Science; English for Writing Research Papers; Science Research Writing for Non-Native Speakers of English; How to Write a Lot; How to Take Smart Notes; Presentation Zen Design; and TED Talks: The Official TED Guide to Public Speaking). The frontmatter provenance block lists the exact principle and claim ids, which resolve into principles/principles.yaml and analysis/claims.jsonl.