What is bear-research-skills?

This repository (README written in Chinese) is a collection of AI-assisted academic research workflows packaged as installable agent skills. The author, who goes by the handle bear, developed these workflows from personal experience doing academic research and released them as reusable tools that plug into AI coding environments like Claude Code or Codex. All of the skills depend on a command-line tool called scimaster-cli, which is a separate academic literature search tool built by the author's team. Every output must trace back to a real search result from that tool. The repository makes this a hard rule: no filling in from memory, no invented references. The individual skills cover common research tasks. bear-support finds real papers that back a given claim, ranked by how directly they support it. bear-counter finds papers that challenge or contradict the same claim, rated by how serious the threat is. bear-map builds a concept knowledge map around a topic using retrieved abstracts, not recalled facts. bear-scoop checks whether a research idea already exists in the literature before you invest time in it. bear-trace traces a research problem back through earlier foundational work to show how a field evolved. The three workflow skills chain these together: bear-review runs support and counter in parallel to give a balanced evidence picture, bear-onboard runs map plus trace to help someone enter an unfamiliar field quickly, and bear-propose runs scoop, support, and counter in sequence to evaluate a proposal idea before it is submitted. Each run produces four outputs: a terminal summary, a structured Markdown report, a self-contained HTML report with dark mode support, and a BibTeX file of all cited papers. The repository is licensed under CC BY-NC-SA 4.0, which permits free use and adaptation for non-commercial purposes with attribution.