The "academic-research-skills" repository by imbad0202 is a comprehensive suite of tools designed to augment academic research workflows using Claude Code, an AI platform. Its primary purpose is to streamline the entire research pipeline—from initial investigation to publication—by automating labor-intensive tasks while keeping the researcher firmly in control. The repository is not intended to write papers autonomously; instead, it assists with reference management, citation formatting, data verification, logical consistency checks, and quality assurance, allowing researchers to focus on critical thinking, methodology selection, and interpretation.
The suite is modular and integrates seamlessly with Claude Code via CLI, VS Code, or JetBrains, with quick installation options. It features a Socratic dialogue mode for planning paper structure, systematic literature reviews (PRISMA), intent detection, and dialogue health monitoring. The system employs a multi-agent architecture: 13 agents for deep research, 12 for academic paper writing, 7 for peer review, and a 10-stage pipeline orchestrator. Each agent specializes in tasks such as style calibration, writing quality checks, LaTeX hardening, visualization, revision coaching, citation conversion, and anti-leakage protocols. The reviewer module uses a multi-perspective approach, including an Editor-in-Chief, dynamic reviewers, and a Devil's Advocate, applying rigorous rubrics and calibration protocols.
A key design principle is the "human-in-the-loop" philosophy. The repository addresses the limitations of fully autonomous AI research systems, such as hallucinated citations, methodology fabrication, and shortcut reliance, by enforcing integrity gates at critical pipeline stages. These gates run a seven-mode checklist to block problematic outputs and offer calibration modes to measure reviewer accuracy against gold standards. The suite incorporates recent research findings on citation reliability and introduces trust-chain frontmatter and locator infrastructure for claim-level audits. It supports claim audits, warning classes for unsupported claims, and calibration thresholds to ensure citation faithfulness.
The repository is highly configurable, supporting multiple languages (English and Traditional Chinese by default, with intent-based activation for others), citation formats (APA 7.0, Chicago, MLA, IEEE, Vancouver), and paper structures (IMRaD, literature review, theoretical analysis, case study, policy brief, conference paper). It provides detailed architecture documentation, performance estimates, and guides for installation and usage. The pipeline is adaptive, with checkpoints requiring user confirmation, mandatory integrity verification, and collaboration quality evaluation.
Showcase artifacts demonstrate real-world outputs, including formatted papers, integrity reports, peer review reports, and post-publication audits. The suite also integrates with the "experiment-agent" companion tool for managing code and human experiments, ensuring validated results before paper writing. Usage is flexible, offering both full pipeline automation and granular skill invocation for research, writing, reviewing, and revision tasks.
Optimizations address AI structural limits such as frame-lock, sycophancy, and intent misdetection. Protocols for Devil's Advocate and Socratic Mentor agents enforce concession thresholds, intent detection, and dialogue health indicators, making AI limitations visible and manageable. Overall, "academic-research-skills" is a robust, open-source toolkit that enhances academic research quality, transparency, and reproducibility by leveraging AI as a collaborative assistant rather than a replacement for human expertise.
