Skip to content

CI: nothing notices when main goes red #43

Description

@castrojo

Found during the CI/CD audit (#42).

main was red on five consecutive pushes (runs 30721325511, 30710115478, 30706187414, 30703310347, 30700620219) and nothing surfaced it: no issue, no alert, no Discord message. The PR gate added in #42 stops most breakage reaching main, but infrastructure failures (runner preemption, cache outage, an upstream FSDK regression) still land there.

Wanted: on a failed Build images run on main, open or update a single tracking issue (or post to the releases webhook), the way projectbluefin/actions factory-health does. One issue that gets reopened, not one per run.

See docs/skills/ci-tooling.md.

Metadata

Metadata

Assignees

No one assigned

    Labels

    3-clanker-queueWork admitted to the agent-maintained queue.

    Type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions