Paper Spine Research Workflow

Problem

Research writing often stalls on engineering-level issues: motivation feels arbitrary, citations are pasted as labels, results drift from claimed contributions, and final delivery reveals that \cite never actually links to the bibliography. Paper Spine is not an abstract polisher. It turns manuscript production into checkable stages where contribution, evidence, reviewer risk, and document format are validated step by step.

How It Works

• Config-first: prefer paper_spine_config.json; if missing, enter the intake flow instead of asking for hand-written JSON.
• Resume-first: run progress_check.py and continue from the first incomplete stage rather than restarting from intake by default.
• Contribution-first: require confirmed_contribution.md, and map each major Results subsection to at least one contribution promise to avoid metric-only writing.
• Reviewer-aware: generate reviewer_audit.md with reviewer value, objection register, and editorial fit before claiming submission readiness.
• Hard delivery: LaTeX must use \cite{key} with bibliography entries, and Word output must pass word_guard.py so hand-typed [1] citations do not silently fail.

Boundaries

Best for rewriting existing papers or building manuscripts from user materials. It does not invent data, p-values, datasets, or experimental claims. If motivation is unconfirmed, the workflow stops at Stage 4. If pandoc or the Word toolchain is unavailable, the run should record BLOCKED instead of claiming completion.