literature-review

Organizes verified Chinese and English sources into evidence-mapped literature review sections.

2|Updated Aug 9, 2026
One-click install
npx skills add https://github.com/DeepJH/doubao-skill-and-info --skill literature-review-deepjh
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: literature-review
Source: https://github.com/DeepJH/doubao-skill-and-info/tree/main/skills/doubao-academic-polish/sub-skills/paper-write-zh/references/literature-review
Command: npx skills add https://github.com/DeepJH/doubao-skill-and-info --skill literature-review-deepjh

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Writing a literature review often leads to fabricated citations, per-paper listing without synthesis, or unverifiable Chinese sources. This Skill enforces a source-first workflow that only writes review paragraphs from verified references, preventing fake citations and empty template prose. ## Core Features & Use Cases - Source Verification Gating: Chinese references must go through browser-task/CNKI verification or manual user backfill; English references use scholar_search metadata plus DOI, publisher, or arXiv pages for stronger evidence. - Evidence-Claim Mapping: Builds a Source ID to claim mapping table before writing, so every literature judgment in the text traces back to a verifiable source with explicit in-text citations (GB/T 7714 by default). - Thematic Synthesis Writing: Organizes reviews by method families, problem conditions, and research gaps instead of author-by-author listing, with quantity targets for technical papers (e.g., 20 English + 10 Chinese sources). - Use Case: A graduate student writing the related work section of a thesis on surface defect detection uses this Skill to verify CNKI and scholar_search sources, build the evidence map, and produce a synthesized review with proper citations. ## Quick Start Help me write the literature review section for my paper on steel surface defect detection, verifying Chinese sources via CNKI and English sources via scholar_search first.

Frequently Asked Questions about literature-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a literature review without fabricating references?

Verify every source before writing: use browser-task to retrieve real CNKI metadata for Chinese literature and scholar_search for English citation metadata. Then build an evidence-claim map linking each source to specific claims before drafting any review paragraph.

How to organize related work by themes instead of listing papers?

Group sources by method family, problem condition, or evidence difference rather than by author. Each thematic paragraph should synthesize two to four sources, state limitations or disagreements, and connect to your research gap, with explicit in-text citations retained.

Can scholar_search results be used as formal references?

Yes, scholar_search metadata supports formal English reference lists, in-text citations, and metadata-level judgments. However, claims about method details, experimental results, or limitations require stronger evidence such as abstracts, DOI pages, publisher pages, or user-provided PDFs.

What happens if CNKI browser-task is unavailable for Chinese sources?

The workflow falls back to manual user backfill using a candidate literature table with title, author, source, year, and link fields. It never substitutes model memory or general search results for unverified Chinese references.

How many references does a technical paper literature review need?

Technical journal or thesis papers target at least 30 sources total: at least 20 English and 10 Chinese. Course papers or proposals target 10 to 15. Gaps must be declared in the delivery status rather than filled with unverified entries.