Variables, contexts and passing data between jobs

Most workflow questions are about moving values around: into steps, between steps, and between jobs.

Environment variables

Use env at three levels. The closest level wins.

Level	Where	Scope
Workflow	top of file	all jobs
Job	under the job	all steps in that job
Step	under the step	that step only

env:
  LEVEL: workflow
jobs:
  test:
    runs-on: ubuntu-latest
    env:
      LEVEL: job
    steps:
      - run: echo "$LEVEL $RUN_NUMBER"
        env:
          RUN_NUMBER: ${{ github.run_number }}

The smallest scope is a step. Read a variable in a shell with $NAME. Read it in YAML expressions with ${{ env.NAME }}.

Contexts and default variables

Expressions use contexts: github, env, vars, secrets, needs, matrix, runner, steps. For example ${{ github.run_number }}. The same data also exists as default environment variables such as GITHUB_RUN_NUMBER, GITHUB_ACTOR and GITHUB_REPOSITORY. GITHUB_ACTOR is the user or app that started the run.

Configuration variables (vars context) hold non-sensitive values and are set at repository, environment or organization level. Use secrets for anything sensitive.

Step outputs and files

Workflow commands write to files. The old set-output command is deprecated.

steps:
  - id: color
    run: echo "action_color=purple" >> "$GITHUB_OUTPUT"
  - run: echo "${{ steps.color.outputs.action_color }}"
  - run: echo "MODE=fast" >> "$GITHUB_ENV"

GITHUB_ENV sets a variable for later steps. GITHUB_OUTPUT sets an output for the step with that id.

Passing data between jobs

Jobs run on separate runners. There are two ways to share data.

• Job outputs: for small strings. Map a step output to jobs.<id>.outputs, then read it as needs.<id>.outputs.<name>.
• Artifacts: for files. Upload with actions/upload-artifact and download in another job with actions/download-artifact.

jobs:
  job1:
    runs-on: ubuntu-latest
    outputs:
      output1: ${{ steps.s.outputs.val }}
    steps:
      - id: s
        run: echo "val=42" >> "$GITHUB_OUTPUT"
  job2:
    needs: job1
    runs-on: ubuntu-latest
    steps:
      - run: echo "${{ needs.job1.outputs.output1 }}"

Jobs run in parallel unless needs creates an order. A matrix such as strategy.matrix.node: [18, 20, 22] creates one job per value.

Exam traps

• variables: and environment: are not valid keys for variables. The key is env, with NAME: value, never NAME = value.
• ${{ GITHUB_RUN_NUMBER }} is invalid. Use ${{ github.run_number }} or $GITHUB_RUN_NUMBER in a shell.
• Cache is not a way to pass data between jobs. Use outputs or artifacts.
• A needs reference only works for jobs listed in needs.
• GITHUB_ENV and GITHUB_OUTPUT are different: environment variable versus step output.

Practice questions

Choose an answer, then review the explanation to check your understanding.

1. Which step correctly sets and uses a custom variable named dbserver?

1. A.variables: dbserver: db1 on a step
2. B.env: dbserver: db1 on a step, then run: echo $dbserver
3. C.environment: dbserver=db1 on a step
4. D.env: - name: dbserver on a step

Show answer

Answer: B.env is a map of key-value pairs. It can be set at workflow, job or step level.

2. A dependent job must read output1 produced by an upstream job named job1. Which expression is correct?

1. A.${{ job1.outputs.output1 }}
2. B.${{ needs.job1.output1 }}
3. C.${{ needs.job1.outputs.output1 }}
4. D.${{ depends.job1.outputs.output1 }}

Show answer

Answer: C.Use the needs context, and the job must list job1 under needs. The path is needs.<job>.outputs.<name>.

3. A step must publish a value so later steps can read it as a step output. Which command does that?

1. A.echo "color=purple" >> $GITHUB_OUTPUT
2. B.echo "color=purple" >> $GITHUB_ENV
3. C.echo "::add-mask::purple"
4. D.echo "::debug::color=purple"

Show answer

Answer: A.$GITHUB_OUTPUT sets step outputs. GITHUB_ENV sets an environment variable for later steps instead.

4. A build must run against Node 18, 20 and 22. What do you use?

1. A.strategy.matrix with a node list
2. B.Three separate workflow files only
3. C.needs with three jobs
4. D.runs-on with three labels

Show answer

Answer: A.A matrix strategy creates one job per value in the list.