Step 5 - Commit and install

Goal: move the reviewed draft out of the disposable .alloy-output/ and into your committed .alloy/, on a branch, so the corpus config is real and reviewable.

Everything so far lived under the gitignored .alloy-output/. Installation is weaving that draft into the .alloy/ frame alloy init --spoke scaffolded. It is all additive, and it belongs on its own branch:

git checkout -b feat/alloy-spoke-corpus

Where each draft file goes

Draft file	Install into
source-entry.yaml	.alloy/alloy.config.yaml, under sources: - replacing the placeholder discovery block
rulepacks/*.yaml and rulepacks/templates/*.md	.alloy/rulepacks/
manifest-categories.yaml	.alloy/corpus/meta/corpus-manifest.yaml, under sources.<name>.categories (add view membership if needed)
units.yaml (if present)	the source's inventory location

A few things to keep right as you weave:

• Keep the source self: true. In a spoke the source is the repo. It carries no repo/ref (those are optional-informational only) and no path/track/drift - its referent is the repo's own HEAD.
• Keep corpusRoot: .alloy/corpus from the scaffold.
• The scaffolded manifest already carries the required scalar fields; you are adding your categories to it, not replacing it.

Verify the installed config

Run the check once more, now against the installed config rather than the draft:

alloy discover           # enumerate the units your installed config declares
alloy status             # coverage, drift, staleness - the whole health report

discover should list exactly the units you settled on in step 4. status reports findings and exits 0 - it never reds anything on a finding, so treat its output as a report, not a gate.

Commit

git add .alloy
git commit -m "feat: install alloy spoke corpus config"

You now have a committed corpus config and an empty corpus. The next step fills it.

Optional: self-updating CI

If you want the spoke to keep itself honest as the code moves, scaffold the spoke CI workflows:

alloy init --spoke --ci     # composable; --ci also works standalone on an existing spoke

This writes .github/workflows/alloy-{detect,propose,validate}.yml and refuses to clobber an existing alloy-*.yml:

• alloy-detect.yml - daily drift detection, rendered into a single pinned issue.
• alloy-propose.yml - a weekly propose run that opens one human-merged PR.
• alloy-validate.yml - corpus validation on every PR that touches .alloy/**.

All three keep the human in the loop - detection and proposal automate; merging never does.

Next: step 6 - Run a campaign.