As AI tools enter daily research work, literature search, idea organization, paper writing and language polishing can each be done by AI. In practice, however, these tasks usually run in isolation: researchers retrieve literature with one tool, hand results to another for organization, re-structure the paper, then check quality with yet another tool - and the information connections between steps are lost along the way.
The open-source Academic Research Skills project on GitHub addresses exactly this. Built for Claude Code, it uses a multi-agent collaboration mechanism to cover literature research, academic writing, paper review and revision, with human confirmation and completeness checks governing the process.
The core design is to organize research tasks into function modules that can be chained together. A unified data-handoff schema lets the research questions, search strategy and evidence material formed in the literature stage flow into paper writing; the finished paper and its research context then flow into review. This reduces repeated input and keeps the information linkage of the whole research process.
The project ships five skills: deep-research (13 agents for deep research and systematic review), academic-paper (writing and revision), academic-paper-reviewer (multi-perspective review), academic-pipeline (unified orchestration of research and writing) and sr-screener (literature screening for systematic reviews). Each can be used alone or orchestrated through academic-pipeline.
For experimental teams publishing in photocatalysis and photoelectrochemistry, the pipeline is most useful at two points: screening large literature sets before designing experiments, and enforcing a review loop before submission. The workflow complements - not replaces - domain judgment.
