Skip to main content

​
Step Conditions

The if: field on a step controls whether it runs, based on the job’s current status.

​
success() (default)

A step with no if: field — or with if: success() — runs only when no previous step in the job has failed: 
- name: Runs normally
  run: echo "all good"

- name: Same behavior, explicit
  if: success()
  run: echo "all good"

​
failure()

Runs only when a previous step has failed: 
- name: This will fail
  run: exit 1

- name: Error handler
  if: failure()
  run: echo "Something went wrong"

​
always()

Runs regardless of prior step status — useful for cleanup: 
- name: Might fail
  run: risky-command

- name: Cleanup
  if: always()
  run: rm -rf /tmp/workdir

​
continue-on-error

When a step has continue-on-error: true, a non-zero exit code does not mark the job as failed. Subsequent steps still see a success() status: 
- name: Optional check
  run: ./optional-lint.sh
  continue-on-error: true

- name: This still runs
  run: echo "Lint result doesn't matter"
Without continue-on-error, a failing step causes all remaining steps (except those with if: failure() or if: always()) to be skipped, and the job is marked as failed.

​
Combining Conditions and continue-on-error

steps:
  - name: Setup
    run: echo "setup"

  - name: Flaky test
    run: ./test.sh
    continue-on-error: true

  - name: Always report
    if: always()
    run: echo "Test completed (may have failed)"

  - name: On success only
    if: success()
    run: echo "Everything passed"
In this example:
  • If test.sh fails, “Always report” runs, and “On success only” also runs (because continue-on-error prevented the job from entering a failed state)
  • If Setup fails, “Flaky test” is skipped, “Always report” runs, “On success only” is skipped

​
Job-Level Failure

When a step fails (without continue-on-error), the job finishes with a failure status. Any downstream jobs that needs: this job are skipped. See Job Dependencies for details on failure propagation across the DAG.