Academic Writing Skills
WenyuChiou/academic-writing-skills
Workflow automatizado para escrita, revisão e submissão de artigos acadêmicos
Ver no GitHubSkill para Claude que automatiza a escrita rigorosa de artigos acadêmicos, incluindo revisão, resposta a reviewers e auditorias pré-submissão. Detecta prosa genérica, verifica consistência entre claims e evidências (figuras, tabelas, estatísticas), e mantém conformidade com regras de periódicos específicos.
Instalação
npx convez-ai WenyuChiou/academic-writing-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
- Revisar seções de manuscritos com estrutura findings-first
- Detectar linguagem vaga, overclaims e inconsistências figura-texto
- Gerar respostas ponto-a-ponto para comentários de reviewers
- Preparar checklists de submissão específicos por periódico
- Auditar claims contra evidências (dados, literatura, código)
Veredito

Skill robusto e bem documentado para pesquisadores automatizarem workflow acadêmico repetitivo. Requer setup inicial por paper (pasta .paper/ com templates), mas economiza tempo significativo em revisão, consistência e conformidade editorial.
README
A tradução para o português deste README ainda está na fila.
academic-writing-skills
A shareable Claude/Codex skill for rigorous academic paper writing, revision, reviewer response, figure-text consistency, and pre-submission audits.
📚 Part of the agentic AI learning roadmap — a 7-stage curated path for building agentic AI, multilingual (zh-TW · zh-CN · English). This skill is referenced in §13 (research workflow skills).
Purpose
This skill helps researchers turn manuscript work into a repeatable workflow:
- Draft sections with findings-first structure.
- Rewrite results and discussion paragraphs with explicit mechanisms.
- Detect GPT-style prose, vague intensifiers, and overclaim language.
- Check whether claims are backed by figures, tables, statistics, code output, or literature.
- Keep numbers, figure panels, captions, and prose consistent.
- Build point-by-point reviewer response tables.
- Prepare journal submission checklists and declaration items.
- Create
.paper/context files so future AI sessions use fewer tokens.
The skill is field-agnostic. Journal-specific rules and paper-specific terminology live inside each paper repository.
Install
Install via the ai-research-skills Claude Code marketplace:
claude plugin marketplace add WenyuChiou/ai-research-skills
claude plugin install academic-writing-skills@ai-research-skills --scope user
Migrating from a previous
git clone? The skill repo was restructured to askills/<name>/SKILL.mdlayout that the marketplace requires. The oldgit clone https://github.com/WenyuChiou/academic-writing-skills ~/.claude/skills/academic-writing-skillspath no longer loads (Claude Code's user-skills loader only scans one level deep). Either uninstall the cloned copy (rm -rf ~/.claude/skills/academic-writing-skills) and run the marketplace install above, or clone the inner subdirectory:git clone -n https://github.com/WenyuChiou/academic-writing-skills /tmp/aws && mv /tmp/aws/skills/academic-writing-skills ~/.claude/skills/academic-writing-skills.
Per-Paper Setup
For each manuscript, create a .paper/ folder in the paper repository:
<paper-repo>/
.paper/
journal_format.md
style_overrides.md
context.md
figure_inventory.md
claim_evidence_ledger.md
reviewer_comments.md
submissions_log.md
Minimum setup:
- Copy
references/journal_format_template.mdto<paper-repo>/.paper/journal_format.md. - Fill in target journal rules from current author guidelines.
- Optionally copy
references/style_overrides_example.mdto<paper-repo>/.paper/style_overrides.md. - For long projects, create the context packet described in
references/paper_context_packet.md.
Skill Layout
academic-writing-skills/
SKILL.md
references/
banned_words.md
claim_evidence_audit.md
figure_conventions.md
journal_format_template.md
paper_context_packet.md
reviewer_response_workflow.md
section_checklists.md
style_overrides_example.md
submission_checklist.md
writing_principles.md
evals/
evals.json
tests/
test_skill_integrity.py
Typical Workflows
Draft or revise a section
The skill loads journal format, paper overrides, writing principles, banned words, and the relevant section checklist. It then rewrites with:
- finding before figure citation,
- mechanism after each result,
- precise terminology,
- no unsupported overclaim,
- panel-specific figure references when needed.
Audit claims and evidence
The skill builds a claim-evidence ledger:
claim -> evidence source -> allowed certainty -> missing check -> revision
This is useful before editing Abstract, Discussion, Conclusion, cover letter, or reviewer response.
Respond to reviewers
The skill converts comments into a response table:
comment -> anchor text -> response -> manuscript change -> evidence
It avoids the common failure mode of saying "clarified" without a visible manuscript change.
Save tokens across AI sessions
For long manuscripts, the skill can maintain .paper/context.md,
.paper/figure_inventory.md, and .paper/claim_evidence_ledger.md. Future
sessions can read those concise files before opening the full manuscript.
What This Skill Is Not
- It is not a Zotero manager.
- It is not an Obsidian or NotebookLM workspace skill.
- It is not a coding-agent delegation skill.
- It is not a generic grammar checker.
- It does not invent scientific assumptions, results, or citations.
Testing
Run integrity tests with:
python -m pytest -q
The tests validate frontmatter, bundled references, eval prompts, and common mojibake/encoding corruption markers.
License
MIT