zchary1106/figurecraft/figurecraft-research-frameworks

FigureCraft Research Frameworks

Turn a study's argument into a figure whose constructs, method, evidence, and outcomes can be verified against the manuscript.

Select the framework grammar

• Conceptual/theoretical: constructs plus typed causal, moderating,

mediating, or associative relationships.

• Methodology: research question → data → method → analysis → evidence.
• Experiment design: factors, controls, treatments, measurements, and

comparisons.

• Technical route: stages, dependencies, feedback, and deliverables.
• Paper method overview: input → modules → outputs plus one representative

detail inset.

• Complex framework matrix: phases across columns, parallel workstreams

across rows, constrained cross-track links, integration hub, and outcome.

Workflow

1. Extract the research question, constructs, hypotheses, inputs, transformations,

evidence, outcomes, and scope from the source text.

1. Separate procedural flow from causal/theoretical relationships.
2. Use diagram.research-framework with framework_stages, stable item IDs,

role-specific cards, concise labels, a central narrative spine, and an optional outer feedback loop. For multiple parallel workstreams, use diagram.framework-matrix with phases, tracks, cross_links, integration, and outcome.

1. Keep one dominant reading direction and no more than four hierarchy levels.
2. Use color for role families: source/input, method, evidence/evaluation,

outcome, and external/context. Use editorial-contrast when stronger color anchors are appropriate, warm-academic for paper-like conceptual work, and midnight-technical only for screen-first dark output.

1. Locate sibling figurecraft/scripts/figure.py, validate, render, and lint.
2. Deliver SVG, FigureSpec, manifest, and Draw.io for manual academic editing.

Edge semantics

• Solid arrow: transformation or supported directional hypothesis.
• Dashed arrow: proposed, optional, feedback, or unverified relationship.
• Plain line: association without direction.
• Edge labels use verbs or hypothesis IDs; never rely on arrow direction alone.

Publication rules

• Every visible construct maps to manuscript text.
• Do not invent hypotheses or causal direction.
• Distinguish data sources, methods, measures, and conclusions by role.
• Put detailed statistics in charts or tables, not tiny framework boxes.
• Split overview and subsystem detail when labels require shrinking.
• Select paper-single/paper-double using the core final-size rules; render-set

retains the complete reference and generates bounded views.

• Put conceptual/source-derived status and key limitations in visible provenance,

not only in a hidden assumptions list.

• Figure title states the framework or claim; the full explanation belongs in

the caption.

Missing constructs, reversed causality, mixed process/causal semantics, unreadable labels, or decorative groups are blocking failures.