The world's most advanced AI-powered research paper creation framework — built for the top 0.0001%.
A comprehensive, skill-based framework that transforms research ideas into papers that pass peer review at the world's most elite venues.
CRES is a comprehensive, skill-based framework that transforms research ideas into papers that pass peer review at Nature, Science, Cell, NEJM, NeurIPS, ICML, or equivalent venues. It provides structured prompts, adversarial review simulations, and systematic workflows that embody the best practices of the world's most elite researchers.
Rather than a single rigid script, CRES is organized as ten modular skills — each activated on demand by a simple command. You compose them like instruments in an orchestra, calling in ideation when you need a sharp research question, methodology when you need a design that pre-empts reviewer critiques, or quality when your draft must survive adversarial review before submission.
The system is powered by ARIA — the Advanced Research Intelligence Architect — a master system prompt that loads a five-step cognitive architecture running silently before every response. The result is a research workflow that is simultaneously systematic and creative — one that respects the rules of rigorous inquiry while leaving room for the judgment of a rigorous mind.
ARIA — Advanced Research Intelligence Architect — is the master system prompt that powers every CRES skill. It loads a six-dimension expertise matrix and runs a five-step cognitive protocol before every response, calibrating output to the highest possible standard.
When you paste MASTER_PROMPT.md as your Claude system prompt, you activate ARIA — a convergence of six elite research capabilities. Every response is then filtered through five cognitive steps designed to find the non-obvious bridge, stress-test against the harshest reviewer, and synthesize across disciplines.
"What does the absolute best version of this look like? Think of the top 3 papers in this space from the last 5 years. What do they share? What do they do that others don't?"
"What is the specific delta between where we are now and that summit? What's the single most important gap to close first?"
"What's the solution that isn't just 'do it better'? What cross-domain insight, methodological innovation, or reframing transforms this problem entirely?"
"I am now the harshest reviewer who has seen 1,000 papers this year and is actively looking for reasons to reject. What do I say? Now: how does the work preempt every one of these critiques?"
"What am I missing by staying within this discipline? What would a physicist / economist / ML researcher / biologist say? What's the cross-domain insight that makes this singular?"
No installation. No dependencies. Just copy, paste, and activate. CRES lives entirely inside your Claude session as a structured system prompt — or as a Claude Project for persistent use.
Paste MASTER_PROMPT.md as your Claude system prompt. This loads the ARIA v3.1 activation system that powers every skill.
MASTER_PROMPT.md
Provide your domain, what you've done so far, your constraints (time, data, resources), and target venue if known.
"I'm working on [TOPIC]"
Issue an activation command to summon the exact skill you need. Skills are modular — call only what you need, in sequence.
"ACTIVATE: IDEATION"
Review each output, provide feedback, then move to the next skill or re-engage the same one for deeper analysis. Best papers are never written in one pass.
iterate → refine
Each workflow chains multiple CRES skills into a coherent, phase-gated process — from the first spark of an idea to post-publication impact. Choose the one that matches your situation.
The definitive A→Z guide to producing a rigorous, high-impact research paper. Integrates all 10 skill modules into eight phase-gated stages with daily task breakdowns, quality gates, and decision points at every transition.
A compressed, high-intensity protocol for time-sensitive research. Trades comprehensive depth for strategic focus while maintaining a non-negotiable quality floor. Includes a decision matrix for when rapid is appropriate vs. risky.
A structured approach to research at the intersection of two or more disciplines. Interdisciplinary work carries extraordinary potential for breakthroughs — and extraordinary risk of failure. This workflow provides the scaffolding to navigate both.
In an era of replication crisis, well-executed replications are among the most valuable scientific contributions — yet among the hardest to publish. This workflow maximizes both rigor and impact across six phases from target selection to publication.
Two timelines, one philosophy: every phase has explicit deliverables, quality gates, and go/no-go criteria. The 28-week track is the gold standard; the 30-day rapid track is for time-sensitive work where data and methods are already in hand.
Define research domain boundaries, scan recent publications, identify 15+ gaps, prioritize by significance/feasibility/novelty, formulate 3–5 candidate hypotheses, select primary hypothesis, and produce a research question document with team sign-off.
Output: Research Question Document + Hypothesis + Novelty Assessment
Systematic search across primary databases (200+ papers), full-text screening (40–70 paper corpus), deep reading in three groups (foundational, recent, cross-disciplinary), citation network analysis, synthesis narrative, and a four-section literature review draft with formatted bibliography.
Output: Literature Review Draft + Citation Network Map + Bibliography
Design experimental protocol, run statistical rigor checks, build reproducibility framework, detect bias, draft pre-registration, and submit ethics application. Pre-registration is non-negotiable — top journals will ask.
Output: Pre-Registered Protocol + Ethics Approval
The longest phase. Collect and clean data, run the analysis engine, generate publication-quality visualizations, and package a reproducible data repository. Every figure must tell a story a reviewer can understand in 30 seconds.
Output: Complete Results + Figures + Data Repository
Draft abstract, introduction, methods, results, discussion, and conclusion — each through its dedicated writing sub-prompt. Format citations in target journal style. Produce a complete first draft, not a polished final.
Output: Complete First Draft
Subject the manuscript to adversarial peer review simulation, logic validation, novelty stress test, and reproducibility audit. Expect REJECT verdicts initially — most papers move to MAJOR REVISION with focused effort on the top 2–3 issues.
Output: Review-Ready Manuscript
Run the venue selection matrix across 5–7 candidate journals scored on scope fit, audience fit, impact match, method affinity, and strategic value. Craft the cover letter and prepare reviewer response templates.
Output: Target Journal + Cover Letter + Response Templates
Final ethics and compliance check, format verification against journal guidelines, submission, and immediate launch of the dissemination plan. Post-publication impact amplification begins the day you submit.
Output: Submitted Manuscript + Dissemination Plan
Compressed gap analysis focused on existing datasets or proven methods. Skip exploratory frontier scanning — go straight to the most tractable, highest-novelty question your data can answer.
Gate: Data ready + method proven + low ethics risk → proceed. Otherwise, abort to full workflow.
Read only the 20–30 papers most directly relevant to your question. Skip broad frontier mapping. Build a tight positioning statement: what gap, why now, why you.
Output: Positioning statement + core citation set
Use a proven, validated method. No new instrument development. Pre-register a minimal protocol. The goal is reproducibility, not methodological innovation.
Output: Pre-registered minimal protocol
Run the analysis. Quantify uncertainty honestly. Generate 3–5 publication-quality figures. No p-hacking — pre-registered analyses only.
Output: Results + figures
Five days, full draft. Abstract last. Cut every sentence that doesn't earn its word count. Aim for short communication length — 2,500–4,000 words.
Output: Complete draft
Run the adversarial review. Fix the top 3 issues only — ignore the rest for the rapid track. Honesty about limitations is mandatory.
Output: Revised draft
Target journals with fast turnaround and short communication formats. PNAS, Nature Communications, journal short-report tracks, or arXiv-first for ML.
Output: Target journal + cover letter
Final polish pass. Format check. Co-author sign-off. This buffer absorbs the inevitable slippage of a compressed timeline.
Output: Submission-ready manuscript
Submit. Immediately post a preprint (if journal permits). Draft a thread, blog post, and outreach list. Day 1 of the citation lifecycle starts now.
Output: Submitted manuscript + preprint + dissemination plan
Every folder serves a distinct phase of the research lifecycle — from the first flicker of an idea to the long tail of post-publication impact.
The repository is organized as a layered system: a master prompt activates the framework, ten skills handle specific research tasks, four workflows chain skills end-to-end, and templates plus evaluators provide format and assessment scaffolding.
claude-research-excellence/ ├── README.md ← You are here ├── MASTER_PROMPT.md ← The ARIA v3.1 activation system ├── SETUP.md ← Setup & best practices │ ├── skills/ │ ├── 01_ideation/ ← Research ideation engine │ ├── 02_literature_command/ ← Deep literature mastery │ ├── 03_methodology_architect/ ← Research design engine │ ├── 04_data_intelligence/ ← Data strategy system │ ├── 05_writing_excellence/ ← Writing mastery system │ ├── 06_quality_fortress/ ← Multi-layer adversarial review │ ├── 07_journal_strategy/ ← Publication targeting system │ ├── 08_ethics_compliance/ ← Research ethics & integrity │ ├── 09_citation_mastery/ ← Citation strategy system │ └── 10_impact_amplifier/ ← Post-publication acceleration │ ├── workflows/ │ ├── FULL_PAPER_WORKFLOW.md ← 28-week A→Z protocol │ ├── RAPID_PUBLICATION.md ← 30-day fast-track │ ├── INTERDISCIPLINARY.md ← Cross-domain innovation │ └── REPLICATION_STUDY.md ← Replication & extension │ ├── templates/ ← Journal & conference formats │ ├── IEEE/ ACM/ Nature_Family/ Science/ │ ├── NeurIPS_ICML_ICLR/ arXiv/ Springer/ │ ├── evaluators/ ← Quantitative rubrics │ ├── NOVELTY_SCORE_RUBRIC.md │ ├── RIGOR_RUBRIC.md │ ├── CLARITY_ANALYZER.md │ └── IMPACT_PREDICTOR.md │ └── examples/ ← Domain-specific examples ├── computer_science/ ├── biomedical/ ├── social_science/ └── interdisciplinary/
Each skill is a self-contained expert you can summon with a single command. Activate one for a focused task, or chain several for an end-to-end workflow.
Transform vague ideas into razor-sharp research questions through field cartography, gap typology, question forging, novelty assessment, and contribution mapping.
ACTIVATE: IDEATIONAchieve genuine mastery & synthesis of a literature — not summaries, but positioned arguments ending with implications for your work.
ACTIVATE: LITERATUREDesign methodology that pre-empts reviewer critiques — experimental design, statistical rigor, reproducibility framework, and bias detection.
ACTIVATE: METHODOLOGYTransform raw results into compelling scientific findings with appropriate uncertainty quantification and publication-quality visualization.
ACTIVATE: DATAProduce rigorous, clear, compelling scientific prose — abstract mastery, introduction craft, methods precision, results power, discussion depth.
ACTIVATE: WRITINGSubject your paper to adversarial review before submission — multi-layer peer review simulation, logic validation, novelty stress test.
ACTIVATE: QUALITYMatch paper to venue & craft winning submission materials — venue selection matrix, cover letter engine, reviewer response mastery.
ACTIVATE: JOURNALEnsure research integrity across all dimensions — bias detection, reproducibility audit, pre-registration, IRB compliance.
ACTIVATE: ETHICSBuild a strategic citation architecture — not exhaustive bibliography, but a network that positions your work and anticipates reviewers.
ACTIVATE: CITATIONMaximize post-publication reach and citations — dissemination plan, preprint strategy, social media, press outreach, follow-up work.
ACTIVATE: IMPACTA look under the hood of Skill 01. Every CRES skill follows this same disciplined structure: phases with explicit quality gates, no skipping, no advancing until the current phase meets its criteria.
The Ideation Engine runs a strict five-phase protocol. Each phase produces concrete artifacts that must clear a quality gate before advancing. This isn't brainstorming — it's a systematic method for locating the highest-novelty, most-feasible, most-impactful question your resources can answer.
The skill integrates forward into Literature Command (02) for gap validation, Methodology Architect (03) for design, and Writing Excellence (05) for articulation. Outputs from Phase 5 directly feed the next skill in the chain.
Map the intellectual terrain. Frontier Scan identifies 5 active research streams and tags each as Advancing, Plateauing, or Retreating. Adjacency Scan finds 3 neighboring fields whose methods could cross-pollinate. Contrarian Probe identifies 3 widely-accepted claims that have never been rigorously challenged.
Classify gaps using the seven-type taxonomy: Knowledge, Method, Theory, Application, Replication, Population, Mechanistic. Score each on Severity, Feasibility, Impact, and Novelty. Compute the Gap Priority Index (GPI) and rank. Only gaps scoring ≥ 3.0 advance.
Transform prioritized gaps into precise, falsifiable research questions. Each question must be answerable in a single sentence, falsifiable, and tied to a specific contribution type. Generate 3–5 candidate questions per gap.
Convert questions into testable hypotheses. Map each to potential methodologies. Evaluate testability and falsifiability. Identify primary hypothesis plus 1–2 backups. The hypothesis must survive the question: "Could a well-designed study plausibly answer this?"
Score novelty across dimensions: conceptual, methodological, empirical, theoretical. At least two dimensions must score high. Produce a contribution map specifying what's genuinely new. Output: the Research Question Document that feeds Skill 02.
Different research situations call for different skill combinations. The matrix below shows the recommended chains for the most common scenarios — drawn from the official SETUP.md guide.
Accept rates, impact factors, and turnaround times for the venues CRES targets. Use this as input to Skill 07 (Journal Strategy) when scoring candidate journals on the five-dimension venue selection matrix.
Distilled from the official SETUP.md. These aren't suggestions — they're the failure modes that the framework exists to prevent.
Common failure patterns and their fixes — drawn from the official SETUP.md. If a phase feels stuck, the answer is almost always to go deeper into a sub-phase, not to skip ahead.
The IDEATION phase is producing incremental questions.
Fix: Focus on the Contrarian Probe and Adjacency Scan sub-phases. Look for assumptions everyone holds and adjacent fields that have solved structurally similar problems.
The LITERATURE phase is producing summaries instead of synthesis.
Fix: Explicitly invoke the Anti-Summary Protocol. Demand that every paragraph ends with an implication for YOUR work, not just a description of prior work.
The QUALITY review is too harsh — producing REJECT verdicts.
Fix: This is working as designed. Use the Revision Priority Order to address the most critical issues first. Most papers move from REJECT to MAJOR REVISION with focused effort on the top 2–3 issues.
I'm not sure which journal to target.
Fix: Run the Journal Strategy venue selection matrix. Score 5–7 candidate journals on all 5 dimensions: Scope Fit, Audience Fit, Impact Match, Method Affinity, Strategic Value. The scores make the right choice clear.
The best research is never produced by a formula — it is produced by a rigorous mind that knows when to follow the rules and when to break them. Use these prompts not as scripts but as frameworks.The CRES Philosophy