The skill
publish-siteas its source filebootstrap/skills/publish-site.mdwrites it, rendered bygen-processes-viz.tsbecause it has no instruction page on this site — do not edit here. All processes
The site: a layer rendered for publication, and its address answering
Skill publish-site: Give a layer a site: its Knowledge Graph rendered for publication at its site address, and the address answering. Four checks — a workflow publishes it, the gh-pages branch exists, Pages serves that branch, the address answers. How the rendering is staged and pushed is the toolset’s own process, not this skill’s.
Named by: Complete initialization, Determine the harness and repositories, Initialize a harness
One abstract step. A layer with a repository is rendered for
publication at its site address, and the address answers. How the
rendering is staged, committed and checked is the toolset’s own subprocess
(its render-to-GitHub-Pages skill and diagram), so this skill stays a list
of checks. The site is how the documents reach their IRIs
(publish-documents, the primary step), not the
goal.
Where the README page lands. At <site address>README.html, and the site
address itself (index.html) is the landing page: by default a plain
redirect to README.html, so a Harness that wants a different front page puts
its own index.html or index.md at its root and changes nothing else. A page generated from the layer’s files says so at its top,
and its footer points to what it was generated from, never to an editor for
the published copy: that copy is replaced on every publish.
The site address is the declaration’s iriBase when it declares one,
otherwise https://<owner>.github.io/<repo>/.
The checks (initialization-steps)
| step | done when |
|---|---|
site:workflow | a workflow commits the rendering onto gh-pages: one under the layer’s own .github/workflows/, or one in the toolset that publishes it from outside. bootstrap’s own workflow writes only its JSON-LD context to gh-pages; its site is published from outside |
site:branch | a gh-pages branch exists. It must exist before Pages can be switched on; where it does not, the step says how to create it (an orphan branch with a placeholder page) |
site:enabled | Pages is on, serving gh-pages at /. Only after site:branch passes for this repository (see Order below). Switched on only by an authenticated forge CLI once the branch exists; otherwise the person gets the one step below |
site:live | the address answers 200, and so does its README page, README.html. 404 after Pages is on usually means the workflow has not run yet. A 403, 407, 5xx or no answer is about the way from here: could not determine, never done |
Order: create gh-pages, then switch Pages on. Never the reverse. site:enabled is not attempted, and not asked of a person, until
site:branch has passed for that repository, checked at the moment of
asking (a read of the branch, not a remembered publish run). A first publish
that is still running, or that failed, has not created the branch. When
several repositories are set up together, the ask lists only those whose
gh-pages exists now; the others are asked later, once their branch exists,
and never ahead of it “for when it appears”. Owner, 2026-11-10: “you need
to create gh-pages first … create gh-pages THEN update settings. you make
this mistake a lot.”
When site:enabled cannot be done here, it goes to the person through
human-agent-discussion, written so it can be
followed without opening anything else:
Open
https://github.com/<owner>/<repo>/settings/pages. Under “Build and deployment”, set Source to Deploy from a branch, branchgh-pages, folder/ (root). It is free for a public repository. Then re-run the workflow that publishes the site from its Actions tab.
If Pages is already on but serves something else, do not switch it yourself: that changes how an existing site is built, so ask.
Never silent: a step that could not be checked is said to be unchecked.
Cost
Pages on a public repository, and the Actions minutes its workflow uses, are free. Nothing else is switched on.