WUBING2023/PaperSpine

What it solves

PaperSpine is designed to prevent common pitfalls in academic writing, such as focusing on sentence-level polishing rather than logical structure, writing without a confirmed core contribution, or presenting results that do not validate the claimed contributions. It ensures that a paper is not only grammatically correct but logically sound and "reviewer-aware," meaning it is structured to withstand the scrutiny of academic peer review.

How it works

The system operates as an orchestration skill for AI agents (supporting hosts like Claude Code, Codex, OpenClaw, and Hermes CLI). It follows a strict 12-stage pipeline that forces the agent to move through specific gates:

1. Research & Planning: The agent learns from target scenarios and high-quality examples, builds a citation support bank, and confirms the paper's motivation and contribution with the user.
2. Execution: It creates a writing rationale matrix and blueprints before drafting. It supports two main workflows: rewriting existing papers or building a new one from raw materials (PDFs, notes, data).
3. Validation: The system uses "hard gates" (Python scripts) to verify that contributions are confirmed, results validate those contributions, and a reviewer audit has been performed.
4. Output: It generates the final manuscript in LaTeX, PDF, and Word formats, along with submission packages (cover letters, highlights) and optional translation packages.

Who it’s for

Researchers, students, and academics writing journal papers, conference papers, technical reports, reviews, or competition entries who want a structured, contribution-driven approach to writing.

Highlights

• Contribution-First Approach: Prevents writing from starting until the core contribution is explicitly confirmed.
• Reviewer-Aware Audit: Generates a reviewer value map and objection registry to simulate peer review before submission.
• Multi-Host Support: Integrates as a skill for Claude Code, Codex, OpenClaw, and Hermes CLI.
• Comprehensive Artifacts: Produces a full audit trail, including a writing rationale matrix and citation support bank, rather than just a final document.
• Humanize Constraints: Includes tiered levels (light, medium, heavy) to reduce AI-generated writing traces.
• Local Reference Indexing: Can index local PDFs, BibTeX, and notes to ground the writing in specific evidence.