The record

A record is a YAML or JSON file. label and description become label and description of the item; every other key is a column of the mapping:

label: Robert David
description: knowledge graph researcher
instanceof: Q5
orcid: "0000-0002-3244-5341"
dblp: "173/3493-1"
linkedInId: "robert-david-39b47692"

a column that is not in the mapping is a problem of the dry run; with --strict nothing is written
a column of the mapping that is not in the record is skipped, unless the mapping gives a fixed value for it
an item value is given as Q id; with valueLookupType in the mapping a name is looked up on Wikidata instead
a list gives one statement per element:

author: [Q115164606, Q55685947, Q56448921]

identifiers with leading zeros or slashes are quoted, so YAML keeps them as text

Dedup first

Never create an item before searching for it. The search shows candidate Q ids with label and description:

ezwd -s "Robert David"

For a paper the DOI is the better key; the Wikidata query service answers whether an item with that DOI (P356) exists. Only when no candidate matches is a record written.

Where the values come from

The record is written from the source of the thing, not from memory: ORCID and dblp for a scholar, the DOI metadata, ceur-spt or dblp for a paper, the volume page for proceedings. The models of the records: Concept:Scholar and Concept:Paper on cr.bitplan.com, Concept:Extension on contexts.bitplan.com.

Procedure

dedup search
record from the source
dry run with --strict, see Dry run
optional sandbox run with --test -w
production write with -w only after a human go, with the acting Wikidata account stated