Research Skills
fangz-cs/research-skills
Skills portáveis para análise estruturada de artigos acadêmicos
Ver no GitHubColeção de Agent Skills portáveis focados em pesquisa acadêmica, com ênfase em leitura crítica de artigos científicos. Funciona nativamente em Cursor, Claude Code e outras plataformas através de um padrão cross-agent agnóstico. Inclui fluxos estruturados para análise bilíngue de papers com notas organizadas por dimensões (motivação, problema, método, trabalhos relacionados, potencial de citação e trabalhos futuros).
Instalação
npx convez-ai fangz-cs/research-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Análise crítica e estruturada de papers acadêmicos
- Geração de notas de leitura bilíngues e organizadas
- Automação de workflows de pesquisa em IDEs de IA
Veredito

Solução bem arquitetada para um problema específico: abstrair a complexidade de ler papers em agentes de IA. O design agnóstico (sem frontmatter agent-específico) é inteligente, mas o catálogo ainda é mínimo (apenas 1 skill ativo). Útil para pesquisadores que usam Cursor ou Claude Code regularmente.
README
A tradução para o português deste README ainda está na fila.
Research Skills
A collection of portable, research-focused Agent Skills (paper reading & more) — Markdown files that teach an AI coding agent how to perform specialized research workflows.
Each skill uses only the cross-agent standard (SKILL.md with name + description frontmatter and plain Markdown), so the same skill works in Cursor, Claude Code, and beyond (Codex CLI, Gemini CLI, Copilot, …). Agent-specific frontmatter is intentionally avoided so nothing gets ignored or breaks.
Skills
| Skill | Description |
|---|---|
reading-papers | Critically read and analyze an academic paper across six dimensions (motivation, problem, method/novelty, related work, citation potential, future work) and produce a structured bilingual reading note. |
Installation
A skill is just a directory containing SKILL.md. Agents discover skills from their own skills folder:
| Agent | Personal (all projects) | Project (per repo) |
|---|---|---|
| Cursor | ~/.cursor/skills/ | .cursor/skills/ |
| Claude Code | ~/.claude/skills/ | .claude/skills/ |
| Other agents | ~/.<agent>/skills/ | .<agent>/skills/ |
Option A — one-click installer (recommended)
git clone https://github.com/fangz-cs/research-skills.git
cd research-skills
./install.sh # install all skills into every detected agent
./install.sh claude # install all skills into Claude Code only
./install.sh cursor reading-papers # install one skill into Cursor only
Usage: ./install.sh [cursor|claude|all] [<skill-name>|all] (both args default to all).
Option B — manual copy
git clone https://github.com/fangz-cs/research-skills.git
# Cursor
cp -r research-skills/skills/reading-papers ~/.cursor/skills/
# Claude Code
cp -r research-skills/skills/reading-papers ~/.claude/skills/
After installing, restart (or reload) your agent so it picks up the new skill.
Usage
Once installed, just ask the agent naturally — for example: "帮我精读这篇论文 ". The agent auto-invokes the relevant skill based on its description.
Repository Layout
.
├── README.md
├── LICENSE
├── install.sh
└── skills/
└── <skill-name>/
└── SKILL.md
Adding a New Skill
- Create a folder
skills/<skill-name>/(the folder name must match thenamein frontmatter). - Add
SKILL.mdwith YAML frontmatter (name,description) and Markdown instructions. - For portability, stick to
name+description+ plain Markdown; keepSKILL.mdunder ~500 lines and move long reference material into sibling files. - Add the skill to the table above.