What problem does it solve? Academic papers need figures and tables that faithfully reflect experimental results, but hand-built charts often hardcode values, mix incompatible comparisons, or lack reproducible sources. This Skill creates evidence-bounded figures, tables, and diagrams directly from authoritative result artifacts. ## Core Features & Use Cases - Evidence-backed plots and tables: Reads raw JSON, JSONL, CSV, or TSV result files so every plotted value traces back to real data, never prose or hardcoded arrays. - Editable diagrams: Produces SVG or TikZ architecture and workflow diagrams using a deterministic renderer instead of generated illustrations. - Rendered visual verification: Inspects each exported PDF, SVG, or raster output for legibility, clipping, caption agreement, and a three-second takeaway test. - Use Case: After finishing an experiment sweep, ask for a multi-panel comparison figure; the Skill reads the canonical result files, drafts a takeaway sentence, generates the plot with its source script under paper/figures/source/, and verifies the rendered output. ## Quick Start Use the paper-figure skill to generate a comparison figure and results table from my experiment result files according to paper/PAPER_PLAN.md.
