iowarp/scientific-method

Attach every literature-derived claim to a specific, checkable citation, and keep the citation set consistent across a document's revisions.

Pick the chart type and encoding from the comparison being made, then execute it with labelled axes and units, visible uncertainty and a self-contained caption, without implying more precision or a stronger effect than the data supports.

Match a research question to a method that can answer it, state the method as a falsifiable procedure with its assumptions and scope, and design the study (variables, controls, sample size, acceptance criterion, confounds) before data is collected.

Turn a set of gathered sources into a defensible state-of-knowledge synthesis with a claim map, contradictions, gaps, and an explicit novelty burden.

Decompose a literature question into search lanes, fetch and read actual sources, and return only what was verified from the fetched text.

Distinguish a real research gap from an under-searched question, and state precisely what evidence would close it.

Work through a self-contained task packet (workspace rules, required outputs and paths, no person available during the run) directly and economically - batch independent calls, read manuals once and search them, verify with bundled checks, keep a facts log, write the required reports once, and confirm every required output before stopping.

Match an analysis approach to the data's actual structure and the question being asked, and state every processing step applied to raw data before reporting a result.

Structure a thesis section, report, or manuscript so every claim traces to its evidence and method, matched to the venue's expected form.

Match a statistical test or distribution fit to the data's actual properties, and report its assumptions and limitations alongside the result.

Track measurement, model, and sampling uncertainty separately and propagate them explicitly through a chain of calculations rather than reporting a single deterministic answer.