Skip to content
José Luis Delgado
Research Engineering Documentation Example MDX

Research Record Workbench

A compact example showing how a technical page can combine a claim, a reproducible workflow, evidence tables, code, and scope.

2026
Portfolio system design
Dark research record workbench with claim, trace, and checks panels

Question

What evidence does the page expose before the reader opens the repository?

This example treats the page as supporting material for a paper. Future research pages use this shape: claim first, evidence second, implementation third, scope visible throughout.

Record contract

SectionPurposeReader question
ClaimState what the page supportsWhat is being asserted?
WorkflowShow the reproducible pathHow would I rerun it?
EvidenceSummarize output and checksWhat did it produce?
LimitsBound the interpretationWhere can this fail?

Minimal workflow

The command block is concrete by design. The template gives future projects an operational path for collection, normalization, and reporting.

pnpm run collect --target tls-boundary-slice
pnpm run normalize --input results/raw
pnpm run report --format static
export type EvidenceRow = {
  target: string;
  observation: string;
  confidence: "low" | "medium" | "high";
  materialUrl: string;
};

export function isReviewable(row: EvidenceRow) {
  return row.confidence !== "low" && row.materialUrl.length > 0;
}

Example evidence

Trace

128

normalized observations

Checks

14

review gates passed

Limits

3

explicit assumptions

Scope

  • Repository links remain the source for implementation detail.
  • Visual summaries carry uncertainty through explicit scope.
  • Missing work appears as open questions and assumptions.

Template use

This prototype sets the shape for future technical pages. A real paper-linked record fills the skeleton with commands, outputs, links, and scope before it becomes part of the public research record.