Purpose

Plan figures and tables for a mathematical modeling contest solution, classifying each visual by its audience and purpose.

This skill determines which figures and tables are needed, what claim each artifact supports, which subquestion it belongs to, which data or result artifact it should use, which of four audience categories it falls into (diagnostic, comparison, paper, appendix), and where it should appear in the paper.

This skill ensures that visual materials serve the modeling logic rather than decoration, and that different audiences (modeler, method selector, paper reader) get the right visuals.

This skill does not fabricate data, generate unsupported figures, write final paper sections, or perform final QA.

B-layer posture — the AI drafts, the modeler owns the claim. Two spans in this plan are interpretive modeling judgments a judge grades, not mechanical bookkeeping: (a) the figure TYPE (1 诊断 / 2 对比 / 3 论文 / 4 附录) — what audience and purpose a figure serves — and (b) the core_claim — the single conclusion a figure is asked to defend in the paper. The AI may DRAFT both, but it must not FINALIZE either on the modeler's behalf:

• The AI proposes a type as [AI-DRAFT — modeler must confirm: <type>]. The human ratifies or changes it.
• For any Type 3 (论文图) figure — every figure that enters the paper — the AI must NOT author the claim. It writes [MODELER INPUT NEEDED: what single claim does this figure defend?] and stops there. A figure's claim is the sentence a judge weighs the figure against; the human authors it.
• These sentinels ([AI-DRAFT, [MODELER INPUT NEEDED) are exactly like the C-layer <<<HUMAN>>> sentinel: a surviving sentinel in a finalized plan is a GATE FAIL. completeness-auditor treats them as "not done". See Rules and Verification.

When to use

Use this skill: