Literature Analysis

Purpose

Generate literature digests, reference lists, and citation analysis reports from PDF or Markdown attachments.

Literature Analysis is the cornerstone of Agentic literature management — every ingested paper should be run through this workflow. It establishes a structured knowledge foundation for each paper, and all advanced features such as citation graphs and Topic Synthesis depend on the outputs of this workflow.

This workflow calls the literature-analysis skill on the Skill-Runner backend to perform structured analysis of academic papers.

Best Practices

• Extract Markdown first: Before running Literature Analysis, it is recommended to use MinerU to convert PDF to Markdown first. The original Markdown significantly improves AI understanding of paper structure.
• Initialize the tag vocabulary first: It is recommended to run Tag Bootstrapper to initialize a controlled tag vocabulary before your first Literature Analysis. This allows the automatic tag regulation in the analysis pipeline to achieve maximum effectiveness.

Use Cases

• Quickly obtain a summary of key content when reading a new paper
• Collect the complete reference list of a paper
• Analyze citation context and citation intent of a paper

Input Constraints

Constraint Type	Description
Input Unit	Attachment
Accepted Types	text/markdown, text/x-markdown, text/plain, application/pdf
Per-parent limit	At most 1 attachment

Trigger Methods

• Directly select a PDF or Markdown attachment
• Select the parent item, and the plugin will automatically expand its first qualifying attachment

Execution Flow

1. Build Request
   └── Upload source file to Skill-Runner
       └── Invoke skill_id: "literature-analysis"

2. Skill-Runner Processing
   └── Parse document content
       └── Generate three outputs:
           ├── digest.md          (Literature Digest)
           ├── references.json    (Reference List)
           └── citation_analysis.json (Citation Analysis)

3. Return Results
   └── Download bundle (zip)
       └── Contains result.json and artifacts/

Execution Mode

Fully automatic, no user intervention required. Simply submit and wait for completion.

Execution Configuration

• execution.mode: auto — Automatic execution, no user intervention required
• skillrunner_mode: auto — Non-interactive mode

Estimated Duration

Scenario	Estimated Time
Standard reference format	6-10 minutes
Non-standard reference format	12-18 minutes

Duration mainly depends on whether the reference format is standard — the more standardized the format (e.g., citations from ScienceDirect, IEEE, and other mainstream journals), the faster AI parsing will be. Paper length has a relatively minor impact.

Outputs

After execution completes, 3 Zotero Notes are created under the parent item:

1. Digest Note

• Type: data-zs-note-kind="digest"
• Content: HTML-rendered literature digest covering research background, methods, results, and conclusions
• Update strategy: Each execution updates the note with the same name (overwrites if it already exists)

About Note Content

The content displayed in the note is rendered from backend data. Directly modifying the note content in Zotero will not change the actual backend data. To edit analysis results, use the Export/Import Notes feature to export, modify, and then re-import.

2. References Note

• Type: data-zs-note-kind="references"
• Content: References HTML table (#, Year, Title, Authors, Source, Locator)
• Update strategy: Each execution updates the note with the same name

3. Citation Analysis Note

• Type: data-zs-note-kind="citation-analysis"
• Content: Citation analysis report including citation context and citation intent classification
• Update strategy: Each execution updates the note with the same name

Parameters

Parameter	Type	Description	Default
language	string	Output language	zh-CN
auto_tag_regulator	boolean	Whether to automatically cascade Tag Regulator after literature analysis. Recommended to enable	true
auto_tag_infer_tag	boolean	When cascading tag regulation, whether to let AI infer new tags (only visible when auto_tag_regulator is enabled)	true

language available values: zh-CN, en-US, ja-JP, ko-KR, de-DE, fr-FR, es-ES, ru-RU. Custom input is also supported.

Model Recommendation

🔴 Models with strong text comprehension are recommended. If the backend supports subagent delegation (e.g., Claude Code, Codex), digest, references, and citation analysis can be processed in parallel, significantly reducing total time.

Dependencies

• Backend: Skill-Runner service
• Backend Configuration: Configure a Skill-Runner type backend in Backend Manager
• Skill: The literature-analysis skill must be deployed on the Skill-Runner

Related Workflows

• Tag Bootstrapper — Initialize a controlled tag vocabulary before your first analysis
• MinerU — Convert PDF to Markdown first for best analysis quality
• Interactive Literature Explainer — Dialogue with AI for deep literature understanding
• Export/Import Notes — Export analysis artifacts for editing, or migrate between Zotero instances
• Tag Regulator — Run tag regulation independently (Literature Analysis can cascade automatically)