research-writing-assistant

Research Writing Assistant

An execution-oriented Skill for undergraduate and graduate thesis writing.

Philosophical Principles

• Process over improvisation — brainstorm first, confirm the structure before writing
• Evidence over claims — citations must be traceable, never fabricate references
• Simplicity over complexity — de-AI-ified writing, reject mechanical expression
• Confirmation over assumption — user confirmation for each chapter, avoid rework

New Architecture (v3.0)

This Skill has been upgraded to a modular skill architecture, supporting multiple platforms:

Platform	Configuration
Claude Code	.claude-plugin/plugin.json
Cursor	.cursor-plugin/plugin.json
Codex	.codex/INSTALL.md
OpenCode	.opencode/INSTALL.md
Gemini CLI	GEMINI.md

Skill List

Core Process Skills

Skill	Path	Description
using-research-writing	skills/using-research-writing/	Entry skill, rules and routing
paper-orchestration	skills/paper-orchestration/	Medium-sized/full-document task orchestration
brainstorming-research	skills/brainstorming-research/	Brainstorming, 7 rounds of Q&A
writing-chapters	skills/writing-chapters/	Chapter writing
latex-output	skills/latex-output/	LaTeX output

Writing Support Skills

Skill	Path	Description
writing-core	skills/writing-core/	Core writing standards
writing-humanities	skills/writing-humanities/	Humanities writing
writing-medical	skills/writing-medical/	Medical writing
writing-law	skills/writing-law/	Legal writing
literature-review	skills/literature-review/	Literature review
evidence-driven-writing	skills/evidence-driven-writing/	Literature-driven introduction/related work
experiment-results-planning	skills/experiment-results-planning/	Experiments, results and mock data planning

Tool Skills

Skill	Path	Description
verification	skills/verification/	Verification mechanism, ensuring completion claims are backed by evidence
figures-python	skills/figures-python/	Python data charts
figures-diagram	skills/figures-diagram/	Flowcharts/architecture diagrams
peer-review	skills/peer-review/	Self-review check
statistical-analysis	skills/statistical-analysis/	Statistical analysis
prompts-collection	skills/prompts-collection/	Prompt collection
environment-setup	skills/environment-setup/	Environment configuration

Red Flags (Stop and Check)

Process-related Red Flags

AI's idea	Correct approach
"The user has explained clearly enough, just start writing"	Must first complete brainstorming-research
"This is just a small revision"	Check whether there is a plan/, if not create one first
"Write a paragraph first to see the effect"	Must first confirm the paper type and chapter structure
"The user is in a hurry, skip the discussion"	The process can be accelerated, but key confirmations cannot be skipped
"This is a simple task, no plan needed"	Any writing task requires a plan record
"I know how to write a paper"	Must write according to the type and structure chosen by the user
"Finish the content first, then deal with the format"	The format is determined during the brainstorming stage
"This chapter's content is simple, no need to confirm"	After each chapter is written, the user must confirm it

Literature-related Red Flags

AI's idea	Correct approach
"I can add some references"	Never fabricate references, they must be traceable
"I remember the content of this skill"	Skills get updated, must re-read the current version
"This citation looks reasonable"	Never write a citation without a source
"The user should know this field"	Do not assume user knowledge; ask clearly before writing

Verification Red Flags (from the verification skill)

AI's thought	Correct approach
"It should be done by now"	Run verification commands to confirm
"The section looks complete"	Run word count and structure checks
"The citation should be real"	Call the CrossRef API or search to verify
"The format should be fine"	Run the format check script
"The search is complete"	Check the number of results and the DOI list
"I am very confident"	Confidence ≠ evidence; run verification
"Skip verification just this once"	No exceptions

Standard workflow

1. Entry routing → call using-research-writing

   • For medium-sized/full-paper tasks, first transfer to paper-orchestration
   • For simple section tasks, continue checking plan/ and section status

2. Medium-sized task orchestration → call paper-orchestration

   • Multi-section, literature-driven, experimental figure, or quality rework tasks must first generate a task package
   • After completion, record the capability-use audit

3. Brainstorming → call brainstorming-research

   • Confirm the paper type (7 types)
   • Confirm the subject area
   • Confirm the title, research background, and methods
   • Confirm the section structure
   • Detect the LaTeX template
   • Create plan/ and chapters/

4. Section writing → call writing-chapters

   • Independent file for each section
   • Two-stage Review (see below)
   • Continue after user confirmation

5. LaTeX output(optional) → call latex-output

   • User provides the template
   • Output .tex files
   • Can be compiled directly

6. Two-stage Review(after each section is completed)

   • Stage one: specification compliance — check whether word count, structure, and citation format meet the requirements
   • Stage two: quality check — check de-AI-ification, language fluency, and academic expression

Module compatibility (backward compatible)

Two-stage Review mechanism

After each section is written, perform a two-stage check:

Stage one: specification compliance check

• Whether the word count meets the requirements
• Whether the section structure is complete
• Whether the citation format is correct
• Whether the subsection titles are clear

Stage two: quality check

• De-AI-ification: no mechanical transition words, no empty emphatic sentences
• Language fluency: no repetitive expressions, no redundancy
• Academic expression: use objective expressions such as "this paper" and "this study"
• Paragraph structure: prefer coherent paragraphs; do not pile up lists

Module compatibility (backward compatible)

Original modules/ The directory is still retained and can be used in the following ways:

Scenario	Module
General Writing	modules/writing-core.md
Humanities Writing	modules/writing-humanities.md
Medical Writing	modules/writing-medical.md
Legal Writing	modules/writing-law.md
Literature Review	modules/literature-review.md
Literature-Driven Writing	skills/evidence-driven-writing/
Paper Workflow Orchestration	skills/paper-orchestration/
Experimental Results Planning	skills/experiment-results-planning/
Translation and Polishing	modules/prompts-collection.md
Self-Review Check	modules/peer-review.md
Python Charts	modules/figures-python.md
Flowcharts	modules/figures-diagram.md
Environment Setup	modules/environment-setup.md
LaTeX Guide	modules/latex-guide.md

Output Specifications

De-AI-ification

• Disable mechanical transition words: first, second, finally, in addition
• Disable empty emphatic sentences: it is worth noting that, it should be pointed out that
• Keep an objective tone, use "this article", "this study"

Format Specifications

• Leave a blank line between paragraphs
• Do not use bold in the main text
• For the main text of a paper, prefer coherent paragraphs and do not pile up lists

Citation Specifications

• Never fabricate references
• Citations must be traceable

Literature and Fact Rules

1. Never fabricate references.
2. English-language literature may be cited after retrieval; for Chinese-language literature, prioritize having the user provide the source.
3. Any citation must be traceable (at least two of author, year, and source must be complete).

Version Information

• Version：3.1.0
• Update Date：2026-05-10
• Maintenance Goals: executable, traceable, process-oriented, multi-platform adaptable

v3.1 Update Contents

1. Philosophical Principles — Clarify core values
2. CSO Optimization — All skill descriptions use the "Use when..." format
3. Enhanced Red Flags — Expanded from 10 items to 12 items
4. Two-Stage Review — Specification compliance + quality check
5. Default Language Rules — Graduation theses default to Chinese; journal submissions follow requirements
6. Structural Templates — Default structural templates for 7 types of papers
7. Interaction Optimization — Let users make fewer choices and more confirmations