imbad0202/academic-research-skills

This is a plugin for Claude Code, the AI coding assistant from Anthropic, that adds a full set of academic research tools to your workflow. Instead of a blank chat window, you get a structured pipeline that walks you through every stage of writing a research paper: planning the paper structure, hunting down sources, drafting sections, checking integrity, running a simulated peer review, revising based on feedback, and producing a finished document. Install takes about 30 seconds through the Claude Code plugin marketplace. The project is built around a clear philosophy: AI should handle the time-consuming background work (finding references, formatting citations, flagging logical inconsistencies, checking for common signs of machine-generated prose) while the human researcher stays in charge of the parts that require genuine judgment, like choosing the research question, interpreting the results, and writing the actual argument. It deliberately avoids full automation, pointing to published research on autonomous AI research systems that still produce hallucinated citations, fabricated results, and methodology errors. The pipeline is broken into stages. Early stages focus on literature review using a 13-agent research team that can query the Semantic Scholar API to verify that cited papers actually exist. Middle stages cover writing, with a feature called Style Calibration that learns from samples of your own past work so the output matches your voice. Later stages include an academic paper reviewer that runs a 7-agent simulated peer review, giving each section a 0 to 100 quality score from multiple perspectives including a Devil's Advocate role meant to stress-test the paper's weakest points. Version 3.3 added several integrity safeguards: a data access level system that tags each tool with how much raw data it is allowed to see (to prevent the reviewer from being contaminated by the ground truth), and an optional artifact reproducibility lockfile that records the configuration used in a run. These features are borrowed from patterns used in AI safety research at Anthropic. Cost for a full 15,000-word paper run is estimated at four to six US dollars in API usage. The output can be formatted as Markdown, DOCX via Pandoc, or PDF via a LaTeX toolchain. A sibling distribution exists for users of the Codex CLI instead of Claude Code. The full README is longer than what was shown.