write-skill

name	write_skill
description	Academic writing skill for generating summaries, reports, literature reviews, and other research documents
version	1.0.0
author	PaperAgent Team

Academic Writing Skill

This skill enables you to generate various types of academic writing, including summaries, literature reviews, research reports, and other scholarly documents. Use this skill when you need to produce well-structured academic content.

Available Tools

You have access to the following writing tools (registered in the Toolkit):

1. generate_summary

Generate a concise summary of given text.

Parameters:

• text (str, required): Text to summarize
• max_length (int, optional): Maximum length in words (default: 500)
• style (str, optional): "academic", "casual", "technical" (default: "academic")
• language (str, optional): Output language - "zh" or "en" (default: "zh")

Returns: Generated summary text.

Example:

generate_summary(
    text=paper_abstract,
    max_length=200,
    style="academic",
    language="zh"
)

2. generate_report_section

Generate a specific section of an academic report.

Parameters:

• topic (str, required): Topic of the section
• content (str, required): Source content or context
• section_type (str, optional): "introduction", "body", "conclusion", "abstract", "methodology", "results", "discussion" (default: "body")
• language (str, optional): Output language (default: "zh")

Returns: Generated section content.

Example:

generate_report_section(
    topic="Transformer Architecture",
    content=research_notes,
    section_type="introduction",
    language="en"
)

3. generate_literature_review

Generate a literature review from a set of papers.

Parameters:

• papers (list[dict], required): List of paper metadata with abstracts
• focus_topic (str, required): Central topic or research question
• structure (str, optional): "thematic", "chronological", "methodological" (default: "thematic")
• length (str, optional): "short", "medium", "long" (default: "medium")
• language (str, optional): Output language (default: "zh")

Returns: Structured literature review with proper citations.

Example:

generate_literature_review(
    papers=paper_list,
    focus_topic="Applications of LLMs in Scientific Research",
    structure="thematic",
    length="long",
    language="zh"
)

4. generate_outline

Generate an outline for a research document.

Parameters:

• topic (str, required): Document topic
• document_type (str, optional): "paper", "thesis", "report", "proposal" (default: "paper")
• depth (int, optional): Outline depth level (default: 3)
• language (str, optional): Output language (default: "zh")

Returns: Hierarchical outline structure.

Example:

generate_outline(
    topic="Deep Learning for Drug Discovery",
    document_type="paper",
    depth=3,
    language="en"
)

5. polish_text

Polish and improve existing text.

Parameters:

• text (str, required): Text to polish
• focus (str, optional): "grammar", "clarity", "conciseness", "academic_tone", "all" (default: "all")
• preserve_meaning (bool, optional): Strictly preserve original meaning (default: True)
• language (str, optional): Text language (default: "zh")

Returns: Polished text with optional change annotations.

Example:

polish_text(
    text=draft_paragraph,
    focus="academic_tone",
    preserve_meaning=True,
    language="en"
)

6. translate_academic

Translate academic text with domain awareness.

Parameters:

• text (str, required): Text to translate
• source_lang (str, required): Source language ("zh" or "en")
• target_lang (str, required): Target language ("zh" or "en")
• domain (str, optional): Academic domain for terminology (e.g., "computer_science", "biology")

Returns: Translated text preserving academic style and terminology.

Example:

translate_academic(
    text=chinese_abstract,
    source_lang="zh",
    target_lang="en",
    domain="computer_science"
)

7. generate_abstract

Generate an abstract for a research paper.

Parameters:

• title (str, required): Paper title
• content (str, required): Paper content or detailed notes
• max_words (int, optional): Maximum words (default: 250)
• structure (str, optional): "structured" (with sections) or "unstructured" (default: "unstructured")
• language (str, optional): Output language (default: "en")

Returns: Generated abstract.

Example:

generate_abstract(
    title="A Novel Approach to Neural Machine Translation",
    content=full_paper_content,
    max_words=300,
    structure="structured",
    language="en"
)

8. generate_related_work

Generate a related work section.

Parameters:

• papers (list[dict], required): Related papers with metadata
• own_contribution (str, required): Description of your own contribution
• comparison_focus (str, optional): What to focus comparison on
• language (str, optional): Output language (default: "zh")

Returns: Related work section with proper positioning.

Example:

generate_related_work(
    papers=related_papers,
    own_contribution="We propose a novel attention mechanism...",
    comparison_focus="efficiency and scalability",
    language="en"
)

Writing Styles

Academic Style

• Formal language
• Third person perspective
• Precise terminology
• Hedging language where appropriate
• Proper citations

Technical Style

• Focus on implementation details
• Code examples where relevant
• Precise specifications
• Step-by-step explanations

Casual Style

• More accessible language
• First/second person allowed
• Simplified explanations
• Engaging narrative

Document Structure Templates

Research Paper

1. Abstract
2. Introduction
3. Related Work
4. Methodology
5. Experiments
6. Results
7. Discussion
8. Conclusion
9. References

Literature Review

1. Introduction
2. Search Methodology
3. Thematic Analysis
   • Theme 1
   • Theme 2
   • ...
4. Synthesis and Gaps
5. Conclusion
6. References

Technical Report

1. Executive Summary
2. Introduction
3. Background
4. Methods
5. Results
6. Discussion
7. Recommendations
8. Appendices

Best Practices

1. Provide rich context: The more context you provide, the better the output
2. Specify structure: Indicate desired organization upfront
3. Review and iterate: Use generated content as a starting point
4. Check citations: Verify all referenced works exist
5. Maintain consistency: Keep terminology consistent throughout
6. Consider audience: Adjust style based on target readers
7. Verify facts: LLM-generated content may contain errors

Output Format

All generated content is in markdown format with:

• Proper heading hierarchy
• Citation placeholders [1], [2], etc.
• Emphasis using bold and italic
• Lists for enumeration
• Block quotes for key points