Research Your Topic With Methodology Guidance3 skills
13-agent research team with guided exploration, systematic reviews, fact-checking, and cross-model verification options.
Research Blueprint
Write Research Papers With Integrity Verification
Imbad0202May 2026
A 4-skill suite orchestrating research, writing, peer review, and publication end-to-end with mandatory integrity gates.
Deep Research & PlanningAcademic WritingIntegrity & Quality VerificationPeer Review & Editorial DecisionCollaborative RevisionFinal Output & FinalizationProcess Reflection & Summary
Install Command
Run this command to deploy the blueprint to your environment.
What problem does this solve?
Runs a full academic research-to-publication pipeline as a connected 10-stage workflow instead of managing research, writing, review, and revision separately. Every stage carries both user decisions and machine-verified integrity checks. The pipeline prevents common failure modes identified in autonomous AI research systems (hallucinated citations, fabricated results, frame-lock, methodology shortcuts) through staged gates and human-in-the-loop verification at critical junctures.
How does it work?
1. Stage 1 (Research): Run deep research on your topic with guidance on methodology, get annotated bibliography and RQ Brief. 2. Stage 2 (Write): Outline your paper with support, then draft with style calibration and anti-leakage checks. 3. Stage 2.5 (Integrity): Machine verifies 7 AI failure modes; you acknowledge the report. 4. Stage 3 (Review): Get 5 peer reviews with editorial decision + revision roadmap. 5. Stages 4-4' (Revise): Incorporate feedback through up to 2 revision loops with Socratic coaching. 6. Stage 4.5 (Final Integrity): Deep verification pass on claims and citations; you confirm ready-to-publish. 7. Stages 5-6 (Finalize): Choose output format and review collaboration quality metrics. Output: peer-reviewed manuscript, integrity reports, process record.
Included Skill Groups
Draft Your Paper With Style & Anti-Leakage Checks3 skills
12-agent writing pipeline with outline approval, style calibration, VLM figure verification, and optional bilingual output.
Verify Claims & Citations at Critical Gates3 skills
Two mandatory integrity gates (2.5 pre-review, 4.5 final) with 7-mode AI failure checklist and optional claim-level audit.
Get Multi-Perspective Review With Editorial Decision4 skills
5-reviewer panel with Devil's Advocate, field-adaptive reviewers, editorial synthesizer, and Sprint Contract protocol.
Revise With Socratic Coaching & Loop Limits3 skills
Up to 2 revision cycles max with optional Socratic coaching; re-review and residual issue resolution.
Format & Publish With Disclosure & Output Verification3 skills
Format selection, AI disclosure statement, and publication-ready output (MD/DOCX/LaTeX/PDF).
Reflect on Collaboration Quality & Process3 skills
Post-pipeline self-reflection, collaboration metrics, and process record generation.
What's the biggest win?
AI-assisted research that surfaces its failure modes explicitly instead of hiding them, so you know which citations to double-check and which claims need external verification before publication.
What should I know technically?
Requires Claude Code v3.7.0+ for plugin packaging; traditional git-clone+symlink install also supported. Pandoc + tectonic optional for DOCX/PDF output (Markdown fallback works). Semantic Scholar, OpenAlex, and Crossref APIs used for reference verification; runs deterministically with cached results when API unavailable. Cross-model verification (GPT-5.5 or Gemini 3.1) supported via ARS_CROSS_MODEL env var for sample integrity checks and independent DA critique.
What should I watch out for?
Stages 2.5 and 4.5 integrity gates cannot be skipped and require user acknowledgment; this adds 20-40 minutes per pipeline run. Material Passport (state file) must persist across session boundaries for mid-entry resume; long-running sessions benefit from checkpoint reset (opt-in). Cite locators require quote/page anchors for strongest claim auditing; title-only citations fall back to advisory. L3 claim audit (opt-in via ARS_CLAIM_AUDIT=1) increases runtime by 15-30% for papers >5k words.
Key Features
10-stage pipeline
Research, writing, integrity gate, peer review, revision cycles, final integrity gate, finalization, and process reflection orchestrated end-to-end.
Mandatory integrity gates
Stages 2.5 and 4.5 run machine-verified checks on 7 AI failure modes (citation hallucination, fabricated results, frame-lock, methodology shortcuts, etc.) before advancing.
Deep research with Socratic guidance
13-agent research team handles systematic reviews, fact-checking, literature analysis, and guided exploration with optional cross-model verification.
Multi-perspective peer review
5-reviewer panel with Devil's Advocate, field-adaptive reviewers, and editorial decision synthesis using Sprint Contract protocol.
Revision coaching with limits
Up to 2 revision loops max with Socratic coaching at decision-heavy moments; prevents infinite revision cycles while maintaining dialogue.
Tools in this Blueprint
About This Blueprint
- License
- CC-BY-NC-4.0
- Industry
- Technology
- Skills
- 1 workflows, 0 sub-skills, 4 standalone