Claude-Research-Paper-OS
CLAUDE.md4 starsChanged 7 months agoApache-2.0View on GitHub
Drop this .claude/ folder into any project directory. Claude Code auto-reads this file on every session start, which activates all agents, skills, and writing rules below. This configuration turns Claude Code into a full academic research and paper-writing system. It provides 6 specialized agents, 47+ skills across 8 skill packs, and an orchestrated pipeline that takes you from literature search to submission-ready manuscript. Claude Code reads this CLAUDE.md on session start. The Skill Auto-Trigger Map below tells Claude which skills…
In this repository6 other files it gives its agents.
- docs/AGENTS.mdAGENTS.md
- .claude/skills/claude-scientific-writer/CLAUDE.mdCLAUDE.md
- .claude/skills/creative-writing-skills/CLAUDE.mdCLAUDE.md
- content-research-writerSkill
- humanizerSkill
- paper-writerSkill
The file
We split this CLAUDE.md into the rules it gives an agent
10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 57
Every section flows like a story.
- Very specific: line 60
Never use : or ; in running text.
- Distinctive: line 63
Never state claims as absolute fact.
- Distinctive: line 75
Every formula needs a plain-English introduction explaining what it computes, followed by a "where" clause defining EVERY symbol used.
- Distinctive: line 78
Every table and figure must appear on the same page as or immediately after its first text reference.
- Distinctive: line 97
Follow venue-specific abstract rules (word count, structure)
- Distinctive: line 106
Always apply Rule 4 (full notation explanation)
- Distinctive: line 110
Run validation script (see Quick Commands below)
- Distinctive: line 115
Always keep near first reference (Rule 5)
- Distinctive: line 161
Run this before finishing any section.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Claude Research Paper OS -- Master Configuration
Drop this
.claude/folder into any project directory. Claude Code auto-reads this file on every session start, which activates all agents, skills, and writing rules below.
What This Does
This configuration turns Claude Code into a full academic research and paper-writing system. It provides 6 specialized agents, 47+ skills across 8 skill packs, and an orchestrated pipeline that takes you from literature search to submission-ready manuscript.
How Auto-Calling Works
Claude Code reads this CLAUDE.md on session start. The Skill Auto-Trigger Map below tells Claude which skills to invoke based on what you ask it to do. You do not need to call skills manually. Just say "write the introduction" or "check citations" and Claude matches your request to the right agents and skills.
For a full autonomous run, say:
Run the full research pipeline for [YOUR TOPIC]Claude will execute the 5-phase pipeline described in docs/PIPELINE.md.
Paper Configuration (Edit These for Your Paper)
# Change these to match your paper
title: "Your Paper Title Here"
authors: "Author One, Author Two"
target_venue: "IEEE Access" # or "NeurIPS", "ACL", "ICML", etc.
format: "IEEEtran" # LaTeX document class
compile_command: >
docker run --rm -v "$(pwd)":/data blang/latex:ubuntu bash -c
"cd /data && pdflatex -interaction=nonstopmode main.tex &&
bibtex main && pdflatex -interaction=nonstopmode main.tex &&
pdflatex -interaction=nonstopmode main.tex"File Structure (Expected)
your-project/
main.tex <- Your paper
references.bib <- Bibliography
figures/ <- Figures
WRITING_RULES.md <- Writing constraints (optional, see templates/)
.claude/
CLAUDE.md <- This file
agents/ <- 6 specialized agents
skills/ <- 8 skill packs, 47+ skillsMANDATORY WRITING RULES
These rules override all default Claude behavior. Every agent and skill follows them.
Rule 0: Story Flow
Every section flows like a story. Before using ANY technical term, explain what it is in plain English. The paper should be accessible from school students to PhD scholars.
Rule 1: No Colons or Semicolons in Prose
Never use : or ; in running text. Use periods, commas, or connecting words instead. Colons and semicolons are allowed in LaTeX commands, tables, and code blocks.
Rule 2: Academic Humility
Never state claims as absolute fact. Use hedged language grounded in evidence.
- Good: "our analysis suggests," "our survey found," "based on our examination"
- Bad: "this proves," "clearly shows," "it is well known that"
Rule 3: No AI Writing Patterns
The following words and phrases are banned in all prose output.
Banned words: landscape (abstract), nascent, pivotal, crucial, testament, showcase, foster, enhance, delve, serves as, stands as, represents a, Additionally, Furthermore, Moreover, Notably, vibrant, tapestry, underscore, intricate, robust, holistic, groundbreaking, cutting-edge, cornerstone, paradigm shift, synergy, streamline, multifaceted, nuanced, comprehensive, leverage, realm, embark, navigate
Banned patterns: em dashes, rule-of-three forcing, synonym cycling, throat-clearing openers ("In the realm of," "It is important to note"), meta-commentary ("This section will discuss"), significance inflation, generic positive conclusions ("the future looks bright")
Rule 4: Formula Explanations
Every formula needs a plain-English introduction explaining what it computes, followed by a "where" clause defining EVERY symbol used.
Rule 5: Tables and Figures Near References
Every table and figure must appear on the same page as or immediately after its first text reference.
SKILL AUTO-TRIGGER MAP
Claude reads this map and automatically invokes the right skills based on what you ask.
When Writing or Rewriting ANY Section
- paper-writing-section -- Section-specific writing guidance
- humanizer -- Anti-AI pattern check on every paragraph
- academic-paper (writing quality check) -- Quality rules from academic-research-skills
- cw-prose-writing -- Narrative flow, sentence rhythm, natural prose style
- cw-story-critique -- Critique flow, pacing, and readability
- claude-scientific-writer -- Scientific writing conventions
- academic-paper-skills (strategist + composer) -- Planning and writing with quality checkpoints
When Writing the Abstract
- paper-writing-section abstract -- Abstract-specific guidance
- Follow venue-specific abstract rules (word count, structure)
When Writing Related Work or Taxonomy
- related-work-writing -- Thematic organization, compare-and-contrast
- survey-generation -- Subsection writing with citation validation
- literature-search -- Verify claims against Semantic Scholar, arXiv, CrossRef
When Working with Formulas
- math-reasoning -- Derivations, notation tables, statistical test selection
- Always apply Rule 4 (full notation explanation)
When Working with Citations
- citation-management -- BibTeX validation, deduplication, auto-fix
- Run validation script (see Quick Commands below)
When Working with Figures or Tables
- figure-generation -- Publication-quality matplotlib figures
- table-generation -- LaTeX booktabs tables
- Always keep near first reference (Rule 5)
When Checking LaTeX
- latex-formatting -- Formatting fixes, pre-submission checks
- paper-compilation -- pdflatex + bibtex pipeline with auto-fix
Before Submission (Review Phase)
- self-review -- 3-persona automated review
- academic-paper-reviewer -- Multi-perspective peer review with 0-100 rubrics
- humanizer -- Final AI pattern removal pass
- citation-management validate -- Final citation check
- paper-assembly -- End-to-end pipeline check
For Deep Research Questions
- deep-research -- 6-phase systematic literature survey
- literature-search -- Multi-source academic search
- novelty-assessment -- Check if claims are novel
QUICK COMMANDS
# Validate all citations
python ~/.claude/skills/agent-research-skills/skills/citation-management/scripts/validate_citations.py --tex main.tex --bib references.bib
# Check LaTeX formatting
python ~/.claude/skills/agent-research-skills/skills/latex-formatting/scripts/latex_checker.py main.tex
# Clean LaTeX artifacts
python ~/.claude/skills/agent-research-skills/skills/latex-formatting/scripts/clean_latex.py .
# Search Semantic Scholar
python ~/.claude/skills/agent-research-skills/skills/deep-research/scripts/search_semantic_scholar.py --query "YOUR QUERY" --max-results 20
# Search CrossRef
python ~/.claude/skills/agent-research-skills/skills/literature-search/scripts/search_crossref.py --query "YOUR QUERY" --rows 10
# Check paper assembly
python ~/.claude/skills/agent-research-skills/skills/paper-assembly/scripts/assembly_checker.py --dir . --verboseWRITING QUALITY CHECKLIST
Run this before finishing any section.
- Zero flagged AI terms (see Rule 3)
- Zero em dashes in prose
- Zero colons or semicolons in prose
- No throat-clearing openers
- No meta-commentary
- No rule-of-three forcing (use 2 or 4 or 5 points if the evidence warrants)
- Paragraph length varies (not all 150-200 words)
- No synonym cycling (one term per concept, stick with it)
- Sentence length varies (no 5+ consecutive same-length sentences)
- All claims hedged with evidence-based language
- All formulas have full notation explanation
- All numbers attributed to sources
- Every technical term explained before first use
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

