GitHub Workflow Automation

Patterns for automating GitHub workflows with AI assistance, inspired by Gemini CLI and modern DevOps practices.

When to Use This Skill

• Automating PR reviews with AI
• Setting up issue triage automation
• Creating GitHub Actions workflows
• Integrating AI into CI/CD pipelines
• Automating Git operations (rebases, cherry-picks)

Full YAML workflow files are in references/workflow_examples.md.

1. Automated PR Review

Create .github/workflows/ai-review.yml triggered on pull_request: [opened, synchronize].

Flow: checkout with fetch-depth: 0 → collect changed files and diff via git diff → send to an LLM (e.g., Claude) → post the response as a PR review comment.

Key elements:

• Permissions: pull-requests: write, contents: read
• Use actions/github-script@v7 to call the AI API and pulls.createReview
• Store the API key in secrets.ANTHROPIC_API_KEY

Review output should follow this structure: Summary → What looks good → Potential Issues → Suggestions → Security Notes.

For focused reviews, filter changed files with grep -E '\.(ts|tsx|js|jsx|py|go)$' before sending to the model.

2. Issue Triage Automation

Auto-label on Open

Create .github/workflows/issue-triage.yml triggered on issues: [opened].

Flow: read issue title & body → call AI with a classification prompt → apply labels (bug, enhancement, question, area labels) → if bug with no repro steps, post a comment asking for details.

Triage prompt returns JSON:

{
  "type": "bug | feature | question | docs | other",
  "severity": "low | medium | high | critical",
  "area": "frontend | backend | api | docs | ci | other",
  "summary": "one-line summary",
  "hasReproSteps": true,
  "suggestedLabels": [],
  "suggestedAssignees": []
}

Stale Issue Management

Use actions/stale@v9 on a daily cron. Mark issues stale after 60 days, close after 14 more. Exempt labels: pinned, security, in-progress.

3. CI/CD Integration

Smart Test Selection

Trigger on pull_request. Determine which test suites to run by inspecting changed paths:

Use a matrix strategy with fromJson() to fan out the test jobs.

Deploy with AI Risk Assessment

On push to main, gather commits since last tag, send to AI for risk analysis (low / medium / high). Fail the pipeline when risk is high to force manual review. Follow with a deployment job gated by the production environment.

Rollback Automation

Create a workflow_dispatch workflow that accepts a reason input. Find the latest stable tag with git tag -l 'v*' --sort=-version:refname | head -1, deploy it, and notify the team via Slack.

4. Git Operations

Automated Rebasing

Listen for /rebase in PR comments via issue_comment: [created]. Checkout the PR, git rebase origin/main, and git push --force-with-lease. Comment the result.

Smart Cherry-Pick

AI-assisted cherry-pick flow:

1. Get commit info and diff against target branch.
2. Ask AI whether conflicts are likely and suggest a resolution strategy.
3. If conflicts expected, create a feature branch, cherry-pick, then resolve each conflict file with AI guidance.
4. Otherwise, cherry-pick directly.

Branch Cleanup

Weekly cron workflow that lists branches not updated in 30+ days (excluding main/develop). Creates a housekeeping issue listing stale branches for team review.

5. On-Demand Assistance

@mention Bot

Listen for @ai-helper in issue/PR comments. Extract the question text, gather context (PR diff or issue body via gh CLI), send both to the AI, and post the response as a comment.

Available Commands

6. Repository Configuration

CODEOWNERS

Map directories and file globs to teams in .github/CODEOWNERS:

* @org/core-team
/src/frontend/ @org/frontend-team
/src/api/ @org/backend-team
/.github/ @org/devops-team
/src/auth/ @org/security-team

Branch Protection

Use repos.updateBranchProtection via actions/github-script to enforce:

• Required status checks (test, lint, ai-review) with strict mode
• At least 1 approving review from code owners; dismiss stale reviews
• Linear history required; no force pushes or deletions

Best Practices

Security

• Store API keys in GitHub Secrets
• Use minimal permissions in workflows
• Validate all inputs; don't expose sensitive data in logs

Performance

• Cache dependencies; use matrix builds for parallel testing
• Skip unnecessary jobs with path filters
• Use self-hosted runners for heavy workloads

Reliability

• Add timeouts to all jobs
• Handle rate limits gracefully with retry logic
• Always have rollback procedures

Resources

• Gemini CLI GitHub Action
• GitHub Actions Documentation
• GitHub REST API
• CODEOWNERS Syntax