Galaxy-Dawn/claude-scholar

Claude Scholar – semi‑automated research assistant

What it is – Claude Scholar is an open‑source extension for Claude Code (and compatible CLIs such as Codex CLI, Kimi Code, OpenCode). It bundles a collection of skills, agents, hooks and rules that let the Claude model help researchers with the repetitive, structure‑heavy parts of a typical computer‑science/AI project: literature hunting, note‑taking, experiment tracking, analysis, and manuscript drafting.

Core idea – Human judgement stays central. Claude Scholar does not try to replace the researcher; it accelerates the workflow around the researcher by providing a reproducible, traceable pipeline:

question → evidence → experiment → analysis → claim → writing

Each stage is stored in a local Obsidian vault (or plain markdown) so the user can inspect, edit, or reject the assistant’s output.

Key components

• Skills – 40+ ready‑to‑use commands (e.g., research‑ideation, literature‑reviewer, results‑analysis, nature‑writing, publication‑chart‑skill).
• Agents – specialised “paper‑miner”, “kaggle‑miner”, etc., that orchestrate multi‑step tasks.
• Hooks & Rules – safety guards (sandbox, permission checks) and style enforcement (coding‑style, experiment‑reproducibility).
• Integrations – optional Zotero MCP for automatic paper import, Obsidian for knowledge‑base navigation, and a plugin marketplace entry for one‑click installation.
• Multi‑CLI support – separate branches (main, codex, kimi, opencode) let you run the same workflow with Claude Code, OpenAI Codex CLI, Kimi Code CLI, or OpenCode.

Installation options

1. Full installer – setup.sh clones the repo, backs up existing Claude config files, writes a manifest, and installs all skills, agents, hooks, and rules. It is incremental and safe to re‑run.
2. Minimal / selective install – copy only the desired skill directories or hook files into ~/.claude/.
3. Marketplace plugin – one‑line command to add the plugin, then manually copy the required rule files.

Typical workflow

1. Ideation – turn a vague research question into concrete hypotheses.
2. Literature – use the Zotero‑linked literature‑reviewer agent to fetch, de‑duplicate, and summarise papers.
3. Note‑taking – convert PDFs into structured markdown notes stored under Sources/Papers.
4. Experiment tracking – results‑analysis generates statistics and figures (leveraging the companion pubfig / pubtab packages).
5. Reporting – results‑report produces a draft report that is automatically synced back to Obsidian.
6. Writing – nature‑writing and nature‑polishing help draft sections, craft arguments, and polish language for a journal submission.

Who benefits

• Computer‑science and AI researchers who move between code, experiments, and papers.
• Graduate students or research engineers needing a reproducible, CLI‑driven workflow.
• Teams that already use Zotero for bibliography and Obsidian for knowledge management.

License – MIT (per the badge in the README).

Status – Actively maintained (last commit shown, recent changelog entries up to June 2026). The project provides multilingual docs (EN, ZH‑CN, JA‑JP) and a growing set of skills.

---

All details above are taken directly from the repository’s README; no additional features have been inferred.