What does this skill do, and when should you use it?
This is a Claude Code skill for academic paper writing, orchestrated by 12 specialized agents through a complete workflow: configuration interview, literature search, architecture design, argumentation, full drafting, citation compliance, bilingual abstract, simulated peer review, and output formatting (LaTeX/DOCX/PDF). It includes style calibration, writing quality checks, and anti-pattern lists to improve writing and avoid common AI traces. Supporting multiple paper types and citation formats, it offers 11 operational modes including plan mode, abstract-only, literature review, and revision coaching.
The skill performs a sequence of operations: conducts a configuration interview (paper type, discipline, journal, citation format, output format), designs a systematic literature search strategy and screens sources, builds paper structure and argument chains, drafts the full text section by section, verifies citations and reference list completeness, generates bilingual abstracts (zh-TW + EN) with keywords, simulates double-blind peer review with five-dimension scoring and revision suggestions, then converts the paper to LaTeX, DOCX (via Pandoc), PDF, or Markdown. If the paper has quantitative results, it can also generate publication-quality figure code in Python (matplotlib) or R (ggplot2).
- A graduate student writing a thesis who needs structured guidance and quality checks
- A researcher with a draft who received reviewer comments and needs a systematic revision plan
- A scholar preparing a journal submission, needing format conversion to the target journal and citation style
- An author requiring bilingual abstracts written independently in both languages, not machine-translated
- A researcher exploring paper structure through Socratic dialogue before writing
How do you install this skill?
- The skill heavily depends on external tools and services (e.g., Pandoc, LaTeX, literature databases) that may be unreachable from mainland-China networks, affecting real-world usability.
- The repository lacks a clear license, and the publisher is unverified; exercise caution regarding copyright and compliance risks.
- Static review cannot verify actual execution; test in a real environment on a small scale before full adoption.
- Shell / CLI
- Network access
- Local filesystem
Pandoc (optional)tectonic (optional)Python 3 (optional)
This skill is part of the GitHub repository Imbad0202/academic-research-skills. Installation method: use Claude Code's plugin marketplace commands:
/plugin marketplace add Imbad0202/academic-research-skills
/plugin install academic-research-skills
or follow other installation methods in the repository's README (e.g., symlink flow).
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/Imbad0202/academic-research-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/academic-paper" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Write a paper on the impact of AI on higher education quality assurance
- Guide me through writing a paper on demographic decline
- Convert to LaTeX
- Parse these reviewer comments
After installation, trigger it with natural language, for example:
"Write a paper on the impact of AI on higher education quality assurance"
"Guide me through writing a paper on demographic decline"
The skill will automatically enter the corresponding mode (e.g., full or plan). Other modes can be triggered with specific phrases like "Convert to LaTeX" or "Parse these reviewer comments".
What are this skill's strengths and limitations?
- End-to-end coverage from planning to submission-ready output, reducing omissions
- Built-in writing quality checks and anti-pattern lists help improve prose quality
- Supports multiple paper structures (IMRaD, literature review, theoretical, etc.) and five citation formats
- Provides 11 modes, allowing isolated use of a single phase (e.g., abstract, literature review)
- Bilingual abstracts are independently written, ensuring quality in both languages
- Relies on LLM generation; accuracy is not guaranteed and human review is still required (especially citations and data)
- Full pipeline can consume significant tokens, leading to higher cost (per performance docs, about $4-6 per full paper)
- Requires user confirmation at each stage, which can be tedious
- Some advanced features (like revision patch) need a Python environment and script support
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Academic Paper Writing Expert Team this page | 0 · Not recommended | ★ 51k | 8d ago | NOASSERTION |
| Multi-Perspective Academic Paper Review Simulator | 52 · Use with care | ★ 51k | 8d ago | NOASSERTION |
| Deep Research Agent Team | 49 · Use with care | ★ 51k | 8d ago | NOASSERTION |
| OfficeCLI Academic Paper Skill | 51 · Use with care | ★ 32k | 5d ago | Apache-2.0 |
| Academic Humanizer | 55 · Use with care | ★ 1.9k | 3mo ago | NOASSERTION |
Compared to the deep-research skill in the same repository, this skill focuses on paper writing (producing a publishable draft), while deep-research focuses on research reports. They can be used together (deep research first, then writing).
How did FollowSkills review this skill?
Evidence shows the skill defines strict role boundaries, write-scope restrictions, checkpoint rules, and IRON RULEs (e.g., user confirmation of config, max two revision rounds), and a Generator-Evaluator contract protocol with clear failure handling and abort tags. These designs imply least privilege and user confirmation, but specifics on data-flow transparency, sensitive-data handling, and external effects (e.g., file writes, external API calls) are lacking. Source attribution is required for citations, but real source verification is not demonstrated. Thus, main risks are visible but permissions, confirmation, isolation, and recovery may be incomplete. Score 0.
Static review cannot execute the skill. The documented happy path is plausible, but no tests, edge cases, or error-handling examples are provided. Though detailed failure handling and fallback strategies exist, there is no execution evidence. Hence capped at 10, but insufficient evidence leads to 0.
The skill defines clear trigger keywords (Chinese, English, Korean) and non-trigger conditions, and distinguishes related skills. It supports Chinese (Traditional and Simplified) and multilingual output. However, the target environment (e.g., mainland-China network reachability) is not addressed. Core functions may rely on external services (e.g., literature search, Pandoc) that might be unreachable from mainland China, so adaptability and environment fit evidence is limited. Score 0.
SKILL.md includes version, last_updated, status, related_skills, known limitations, and routing discipline. Quick start and execution flow are provided. But install/dependency notes (e.g., requires Pandoc, LaTeX), parameter naming stability (unclear), examples and FAQs, changelog, maintenance responsibility (not declared), and license information (repository lacks explicit license, marked NOASSERTION) are missing. Thus information architecture and governance are incomplete. Score 0.
Static review cannot verify whether the skill actually completes user tasks. The docs claim outputs are directly usable (e.g., LaTeX/DOCX/PDF), but no actual output examples or evidence are provided. Marginal value and cost/benefit ratio are not evaluated. Hence capped at 7, but insufficient evidence leads to 0.
No test suites or CI workflows are found in the repository. All claims rely on self-documentation, with no third-party execution evidence. Therefore verifiability is low. Score 0.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →