Academic Writing Skills

Develop and review a manuscript as one connected evidence system—from argument architecture and the extended outline to drafting, full-manuscript review, revision, and final submission.

Built on the open Agent Skills format. Designed for Claude Code, ChatGPT, Codex, OpenCode, Hermes Agent, and other skills-compatible agents. 繁體中文

From research architecture to submission

A general-purpose AI can improve one paragraph while missing what changed elsewhere in the paper. These skills manage the complete manuscript lifecycle:

Research framing → Argument architecture → Extended outline → Evidence-led drafting → Bidirectional alignment → Full-manuscript review → Revision → Submission verification

Top-down review follows the gap or purpose through the questions, methods, evidence, and conclusions. Bottom-up review starts from source evidence and tests whether each result, interpretation, contribution, and summary claim is actually supported.

Two skills, one manuscript system

paper-review is review-only by default. After choosing which comments to apply, use academic-writing-skills to make the revisions and propagate their effects through the manuscript system.

Install and use

The prompts below are platform-neutral: name the skill in ordinary language. Compatible agents may also activate it automatically or offer their own skill selector or shortcut.

For Claude Code:

claude plugin marketplace add WenyuChiou/ai-research-skills
claude plugin install academic-writing-skills@ai-research-skills --scope user

One installation provides both skills. To update later:

claude plugin update academic-writing-skills@ai-research-skills

For Codex, OpenCode, Hermes Agent, or another Agent Skills client, clone this repository and copy or link both folders under skills/ into a directory the client scans. A shared .agents/skills/ directory can serve Codex and OpenCode; Hermes Agent can also scan that directory when configured as an external source.

Try it

Attach the exact active manuscript, section, or outline. Add figures, tables, supplements, reviewer comments, prior versions, or venue guidance when they control the task.

Develop an extended outline

Use the academic-writing-skills skill to develop an extended outline from the
attached materials. Establish the gap, research questions, planned evidence,
intended contribution, and nonclaims. Give every planned paragraph a function,
claim, authorized evidence, inference limit, and bridge. Do not draft the full
manuscript yet.

Review the complete manuscript

Use the paper-review skill to conduct a four-pass top-to-bottom review of the
attached manuscript and supplement: argument and structure; evidence and scope;
scholarly writing and flow; and delivery integrity. Rank issues by scientific
and reproducibility risk, anchor each comment to the files, and do not edit
them.

What it protects

• Question-to-evidence alignment from research architecture through the final summaries.
• Scientific meaning, locked decisions, claim scope, and explicit nonclaims.
• Consistency across the manuscript, Abstract, figures, tables, supplements, reviewer responses, metadata, and submission files.
• Terminology, necessary versus avoidable repetition, paragraph flow, and stock phrasing without cosmetic synonym swapping.
• Relevant technical risks in equations and derived displays, surveys and psychometrics/SEM, simulations and AI/LLM studies, water and flood models, and revision rounds.

Missing results, citations, assumptions, and reviewer preferences are reported as limitations; they are never invented.

Learn more

• Full usage guide — task inputs, more platform-neutral prompts, review stages, technical modules, and long-running project support.
• Release history

Part of the agentic AI learning roadmap.

License

MIT