Academic Paper Team Collaboration System

Problem addressed

Academic paper work split among “experts” often becomes a patchwork: literature, simulation, writing, and review drift apart, state stays inconsistent, placeholders remain, and figures or data cannot be traced back to claims. academic-paper-team targets that multi-agent workflow problem: keeping a reliable Paper State across a long session and turning drafting into verifiable stages.

How it works

The skill uses a Controller that does not write the manuscript directly. It coordinates state, task dispatch, conflict merging, and user decisions. The workflow is built around 4 Generic Experts, 3+1 Ledgers, 5 Gates, Sprint Contract, and verification discipline.

• config.json stores paper type, target journals, citation style, and other routing parameters.
• research, review, case_study, and methodology use different stage routes, with simulation and gate checks adjusted accordingly.
• Each stage starts with a Sprint Contract and is accepted item by item at gate review, rejecting “looks correct” style sign-off.
• academic-reviewer acts as an independent evaluator and must cite paragraphs, figure/table IDs, severity, and fixes; if unavailable, a degraded path uses academic-lit-expert.
• Intermediate artifacts are isolated in /tmp/paper-workspace-{timestamp}/, while the source skill files remain read-only.

Boundaries

This is best suited to papers that need multi-turn collaboration, evidence checks, journal formatting, and final .docx generation. It is heavier than needed for one-shot summaries, simple layout, or short text without literature and method chains. Users should clearly record target journals, citation style, and unresolved placeholders, otherwise Gate 4 may stall on evidence or formatting consistency.