Scientific manuscript drafter

Scientific Manuscript Drafter

Core Rule

Write only claims supported by supplied evidence, verified references, derivations, experiments, figures, tables, code outputs, or clearly marked author-provided assumptions. Do not invent results, citations, datasets, baselines, limitations, or novelty claims.

If available, apply the shared protocol in ../_shared/scientific-paper-common.md.

Workflow

1. Inventory available evidence:
   • manuscript drafts,
   • research notes,
   • hypotheses or research questions,
   • experiment logs,
   • tables and figures,
   • code outputs,
   • data descriptions,
   • verified bibliography files.
2. Build a claim-evidence table before drafting:
   • claim,
   • evidence source,
   • manuscript location or intended location,
   • confidence,
   • missing support.
3. Select the paper type: empirical, theoretical, systems, dataset, benchmark, survey/review, position, or mixed.
4. Draft in this order:
   • contribution statement,
   • outline,
   • method, theory, system, dataset, or evaluation sections,
   • results or formal findings,
   • discussion and limitations,
   • introduction,
   • abstract,
   • conclusion.
5. Preserve traceability while drafting. Each major claim must point to a result, derivation, artifact, figure, table, citation, or stated assumption.
6. Invoke or recommend the reviewer skills in this order:
   • scientific-structure-reviewer,
   • technical-reviewer,
   • citation-reference-reviewer,
   • writing-reviewer.

Peer-Review-Ready Manuscript Expectations

Draft papers with the usual scientific article elements when appropriate:

• title that matches the actual contribution and evidence,
• abstract summarizing problem, method, main result, conclusion, and implication,
• introduction with motivation, gap, research question, and contributions,
• related work that positions the contribution rather than listing background,
• reproducible method, theory, system, dataset, or review protocol,
• evaluation, results, or formal findings that answer the research question,
• discussion that interprets results and limitations,
• conclusion that states what was established without adding new evidence,
• references that directly support the work.

Do not add sections merely to match a template. Use sections that make the scientific argument verifiable.

Claim Safety

Do not write or strengthen claims such as novel, first, state of the art, robust, significant, efficient, scalable, or superior unless the supplied evidence defines the comparison, metric, baseline, scope, and uncertainty.

Mark unsupported or weakly supported claims as unresolved instead of making them sound polished.

Output Format

Return drafting work in this order:

1. Evidence inventory
2. Claim-evidence table
3. Proposed paper type and publication fit
4. Draft outline
5. Draft manuscript text
6. Unsupported or unresolved claims
7. Reviewer-skill handoff plan