Push to origin from GitHub action

actions/checkout@v2

Version 2 of checkout resolves the detached HEAD state issue and simplifies pushing to origin.

name: Push commit
on: push
jobs:
  report:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v2
      - name: Create report file
        run: date +%s > report.txt
      - name: Commit report
        run: |
          git config --global user.name 'Your Name'
          git config --global user.email '[email protected]'
          git commit -am "Automated report"
          git push

If you need the push event to trigger other workflows, use a repo scoped Personal Access Token.

      - uses: actions/checkout@v2
        with:
          token: ${{ secrets.PAT }}

actions/checkout@v1 (original answer)

To add some further detail to the excellent answer by @rmunn. The problem is that the actions/checkout@v1 action leaves the git repository in a detached HEAD state. See this issue about it for more detailed information: https://github.com/actions/checkout/issues/6

Here is a complete example to demonstrate how to get the checked out repository to a usable state and push to the remote.

name: Push commit
on: push
jobs:
  report:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v1
      - name: Create report file
        run: date +%s > report.txt
      - name: Commit report
        run: |
          git config --global user.name 'Your Name'
          git config --global user.email '[email protected]'
          git remote set-url origin https://x-access-token:${{ secrets.GITHUB_TOKEN }}@github.com/$GITHUB_REPOSITORY
          git checkout "${GITHUB_REF:11}"
          git commit -am "Automated report"
          git push

To include untracked (new) files change the workflow to use the following.

          git add -A
          git commit -m "Automated report"

The above workflow should work for the majority of events. For on: pull_request workflows the merging branch (GITHUB_HEAD_REF) should be checked out to replace the default merge commit.

Important: If you have other pull request checks besides the following workflow then you must use a Personal Access Token instead of the default GITHUB_TOKEN. This is due to a deliberate limitation imposed by GitHub Actions that events raised by a workflow (such as push) cannot trigger further workflow runs. This is to prevent accidental "infinite loop" situations, and as an anti-abuse measure. Using a repo scoped Personal Access Token is an approved workaround. See this GitHub issue for further detail on the workaround.

name: Push commit on pull request
on: pull_request
jobs:
  report:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v1
        with:
          ref: ${{ github.head_ref }}
      - name: Create report file
        run: date +%s > report.txt
      - name: Commit report
        run: |
          git config --global user.name 'Your Name'
          git config --global user.email '[email protected]'
          git remote set-url origin https://x-access-token:${{ secrets.GITHUB_TOKEN }}@github.com/${{ github.repository }}
          git commit -am "Automated report"
          git push

For further examples of push to origin during an on: pull_request workflow see this blog post, GitHub Actions: How to Automate Code Formatting in Pull Requests.

You can use secrets.GITHUB_TOKEN as a password on your repository URL. So you might add this before your git push line:

git remote set-url --push origin https://your_username:[email protected]/your/repo

This assumes that you're already passing in the GITHUB_TOKEN secret as an environment variable to your script. If you aren't, then add:

env:
  GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

to your workflow step.

All the provided answers are correct. But I would like to add a bit more to these answers. There can be situations where there are no changes in files. So in that situation, when trying to execute the git commit command it returns an error that fails the GitHub workflow (failed workflows block the PR merge). So we have to check whether there are any changes in files before commit.

name: Config Generator
on:
  pull_request:
    branches: [ main ]

jobs:
  config-generator:
    runs-on: ubuntu-latest
    steps: 
      - uses: actions/checkout@v2
        with:
          ref: ${{ github.head_ref }}

      - name: Install jsonnet
        run: sudo apt install -y jsonnet

      - name: Generate probe configs
        run: python3 generate-configs.py

      - name: Check for modified files
        id: git-check
        run: echo ::set-output name=modified::$(if [ -n "$(git status --porcelain)" ]; then echo "true"; else echo "false"; fi)

      - name: Update changes in GitHub repository
        if: steps.git-check.outputs.modified == 'true'
        run:  |
          git config --global user.name 'Your Name' 
          git config --global user.email 'Your Email for GitHub'
          git add -A
          git commit -m '[automated commit] add configs generated using jsonnet & GitHub workflow'
          git push

The above workflow is an example for creating configuration files using jsonnet and GitHub workflow. It first installs jsonnet and runs a python script that creates the config files using jsonnet templates. As mentioned above there can be situation where there are no file changes. So it uses git status command to check whether there are any file changes. (Instead of git status, git diff can also be used. But it wont show untracked files). Rest of the git commands run only if there are file changes.

Also note that I had to use ref: ${{ github.head_ref }} to checkout to source branch even though I used checkout@v2 (still detached head problem exists in my case)