Actions And Runs

Turn automation into accountable work

Actions define reusable workflow logic. Runs, jobs, and artifacts show what happened when that logic was executed. Together they turn automation into something operators can inspect instead of a black box.

Follow the execution chain

flowchart LR
    Action[Action] --> Run[Run]
    Run --> Job[Job progress]
    Run --> Artifact[Artifacts]
    Artifact --> Review[Human review]

The operating model is simple:

• the action defines the reusable behavior
• the run records a specific execution attempt
• the job shows queue-backed progress when work is asynchronous
• artifacts hold the output the team may need to inspect, review, or reuse

Use it when the workflow repeats

This model matters when:

• the same operational logic should be reused across records or teams
• a run may take time, fail, or need retry
• the resulting output has to be reviewed before it is trusted
• AI-assisted execution is part of a larger operational workflow

Keep review in the loop

The important boundary is that a successful run is not automatically an approved result. Retry is not a substitute for fixing weak inputs, and artifacts are part of the operational record rather than disposable byproducts.