peacefulj/SSCI-SCI-skills
Learning ResourcesA paper-writing assistant that applies a 393-rule academic writing library through lexical checks and LLM judgment, from topic selection to revision.
This repo is an agent skill pack · 1 skillsSkill list and install →What it does
PaperMate packages a 393-rule academic writing and review knowledge base into a standalone web app. The README positions it for students and researchers working on SSCI/SCI-style English papers: it covers topic selection, research-design review, drafting guidance, citation formatting (IEEE/APA/GB7714/footnotes), pre-submission checks and rebuttal structure. Of the 393 rules, 95 can be checked lexically by the bundled skill scripts; the remaining 298 require semantic judgment, which the backend sends to DeepSeek, Anthropic or Lumda AI in batches README.
A separate radar add-on queries OpenAlex publication trends and produces three scored candidate topics. A seven-step one-shot pipeline can generate an English working draft with [TBD: ...] data placeholders and [CITE: ...] citation placeholders instead of fabricated numbers or references.
Quick start
The README offers three routes. For Windows users the recommended path is the installer (no Python needed): run PaperMate-Setup-1.3.0.exe, complete the environment checks, and the browser opens once the service starts. Program files go under %LOCALAPPDATA%\Programs\PaperMate; uploads, drafts and the API key stay under %LOCALAPPDATA%\PaperMate README.
To run from source:
cd academic-writing-agent
python env_check.py
cp .env.example .env # then fill DEEPSEEK_API_KEY
./start.sh # Windows: start.bat
Then open http://127.0.0.1:8000 (the port shifts to 8001/8002 if taken). Python 3.10+ is required README. To run it on a server with password auth and persisted state, use the Docker or systemd route in the deployment guide.
After launch, configure one of the three providers — DeepSeek, Anthropic or Lumda AI — in Settings and click "Test connection" to check reachability.
Constraints
- The rule library targets English IMRaD papers. Purely theoretical/mathematical papers (definition–theorem–proof) are unsupported; basic-science and clinical tracks are lower-confidence, while quantitative empirical and qualitative work is the most validated README.
- Lemma-level checks cover only 95 of the 393 rules; the rest rely on LLM judgment and may err. Major conclusions need human review.
- One-shot drafts are working drafts only. Every generated text is AI-generated content that may carry watermarks; exports force an
ai_generated: truewarning header. The intended use is as a structural skeleton that the author rewrites. - Uploaded data files: only structure (column names, types, missing rates, summaries) is sent to the model; raw rows are not stored server-side.
- The README lists an MIT license plus usage notes, but the repository metadata reports NOASSERTION, so the exact license terms should be checked in the repo before redistribution.
HotRepo's take
This repo appeared on HotRepo's new-project board for Python on 2026-11-20, when it added 5 stars in a roughly two-hour window to reach 106. There is no all-language daily-rank record and no Hacker News discussion, so the data supports only a single short-window observation. Its distinguishing detail is architectural rather than promotional: it keeps the original rule scripts untouched, re-imports them for lexical checks, and mounts the topic radar as a removable add-on whose verify command hash-checks the base rules. The project's own capability-boundaries section, which admits that weak tracks and LLM misjudgment exist, is the most useful part of the README for deciding whether it fits a given paper.
Written by AI from the project's README, updated . The project's own docs are the source of truth.
You might also like
- Imbad0202/academic-research-skillsA suite of Claude Code skills that guides researchers through literature search, drafting, peer review and revision with human checkpoints.
- Shubhamsaboo/awesome-llm-appsA collection of runnable AI agent, RAG and agent-skill examples for developers learning to build LLM applications.