Much like Google Structured Data we can place a simple descriptor with a well-known name like kg.jsonld into source code repos, crawl the repos, build a part of a knowledge graph and federate across other parts of the knowledge graph built with, for example, Confluence content using Taxonomies for Confluence add-on.

Alternatively, contents of kg.jsonld can simply be pushed via SPARQL 1.1 Graph Store HTTP Protocol to a graph store via Github Actions or Bitbucket Pipelines.

Another alternative is Bitbucket webhooks, so it is possible to build a single knowledge graph ecosystem around Atlassian tool set.

Only two terms are used to describe resources - dcterms:type and dcterms:subject from Dublin Core Metadata Terms. Type can be a concept from SDLC Document Types Taxonomy or from an other scheme that may be organisation-specific. Subject can be a concept referencing a system, component, service, or anything else from an organization-specific scheme.

Simplest case of specifying type and subject of a single README file and using remote context hosted in this repo:

{
  "@context": "https://raw.githubusercontent.com/cadmiumkitty/kg-file/main/kgcontext.jsonld",
  "@graph": [
    {
      "@id": "README.md",
      "type": "https://dalstonsemantics.com/taxonomy/sdlc/readme",
      "subject": "https://dalstonsemantics.com/service/thingymabob"
    }
  ]
}

A case can be made for something even simpler that will be translated by the crawler/action/pipeline/webhook processor:

README.md:
  type: https://dalstonsemantics.com/taxonomy/sdlc/readme
  subject: https://dalstonsemantics.com/service/thingymabob