Gate CI on fleet drift

gh-aw-fleet status and gh-aw-fleet overview exit non-zero when a repository has drifted from fleet.json. That makes either one a drop-in CI gate: run it as a step and let the exit code fail the job.

Use `status` for a fast, clone-free gate

status diffs desired versus actual workflow refs over the API without cloning anything, so it is the cheapest gate:

gh-aw-fleet status

Exit 0 — every tracked repo is aligned.
Exit 1 — at least one repo has drifted (or errored).

Drop it straight into a workflow step; a non-zero exit fails the job:

- name: Fail on fleet drift
  run: gh-aw-fleet status

Inspect drift in JSON

To report which repos drifted rather than only failing, read the JSON envelope:

gh-aw-fleet status -o json \
  | jq '.result.repos | map(select(.drift_state == "drifted")) | length'

Scope the check to a single repo by passing it as an argument (status accepts at most one; omit it to check the whole fleet):

gh-aw-fleet status you/your-repo

Use `overview` when you also want health and cost

overview shares the same drift-only exit contract but additionally reports run health, no-op rate, and AI-credit spend. Its window defaults to the trailing 7 days:

gh-aw-fleet overview

Exit 0 — every in-scope repo is aligned.
Exit 1 — any in-scope repo is drifted or errored.

Run failures stay advisory: a fully aligned fleet with failing agentic runs still exits 0, so overview gates on drift, not on flaky runs. It reuses the gh aw logs fan-out, so it is slower than status and can take minutes on large fleets — prefer status when you only need the drift gate.

Gate CI on fleet drift

Use status for a fast, clone-free gate

Inspect drift in JSON

Use overview when you also want health and cost

See also

Use `status` for a fast, clone-free gate

Use `overview` when you also want health and cost