🔧 GitHub Workflow Automation

Automate GitHub workflows with AI-powered assistance using GitHub Actions, inspired by Gemini CLI and modern DevOps practices.

Overview

This skill provides patterns for integrating AI into GitHub workflows to automate code reviews, issue management, CI/CD pipelines, and Git operations. It focuses on practical, production-ready GitHub Actions configurations that leverage Claude/AI for intelligent automation.

Core Capabilities

• AI-Powered PR Reviews: Automated code review with structured feedback
• Issue Triage: Auto-label, categorize, and manage incoming issues
• CI/CD Integration: Smart test selection, deployment validation, rollback automation
• Git Operations: Automated rebase, cherry-pick, and branch cleanup
• On-Demand Assistance: @mention bots for interactive help in PRs and issues

Detailed Workflow Guides

Choose the appropriate reference based on your automation needs:

PR Review Automation

Use when: Setting up automated code review workflows, implementing review patterns, creating focused filters.

Covers:

• Complete GitHub Actions workflow for AI-powered PR reviews
• Structured review comment patterns
• File type filtering and context inclusion
• Security best practices

Issue Triage

Use when: Automating issue labeling, managing stale issues, implementing first-response templates.

Covers:

• Auto-labeling based on issue analysis
• Triage prompt patterns
• Stale issue management workflows
• First-response automation

CI/CD Integration

Use when: Building smart CI/CD pipelines, implementing deployment validation, creating rollback automation.

Covers:

• Smart test selection based on file changes
• AI-powered deployment risk assessment
• Automated rollback workflows with notifications

Git Operations

Use when: Automating Git workflows, handling rebase/cherry-pick operations, managing branch lifecycle.

Covers:

• Comment-triggered auto-rebase
• AI-assisted cherry-pick with conflict resolution
• Weekly branch cleanup automation

Repository Configuration

Use when: Setting up CODEOWNERS, configuring branch protection, implementing @mention bots.

Covers:

• CODEOWNERS file patterns
• Branch protection via GitHub API
• @mention bot for on-demand assistance
• Available commands reference

Quick Start

1. Choose your automation type from the list above

2. Read the appropriate reference guide with complete workflow examples

3. Adapt to your repository:

• Replace placeholders (repo names, team names)
• Configure secrets (AI API keys)
• Adjust thresholds and rules

4. Test in staging before production deployment

Best Practices

Security

• Store AI API keys in GitHub Secrets (ANTHROPIC_API_KEY, etc.)
• Use minimal permissions in workflows (contents: read, pull-requests: write)
• Validate all inputs before processing
• Never expose sensitive data in workflow logs

Performance

• Use path filters to skip unnecessary workflows
• Implement smart test selection to reduce CI time
• Cache dependencies appropriately
• Consider self-hosted runners for heavy workloads

Reliability

• Add timeouts to all jobs (default: 360 min may be too long)
• Handle API rate limits gracefully
• Implement retry logic for transient failures
• Maintain rollback procedures for all automations

Common Patterns

Trigger on PR events:

on:
  pull_request:
    types: [opened, synchronize, reopened]

Use outputs between jobs:

jobs:
  analyze:
    outputs:
      result: ${{ steps.step-id.outputs.value }}
  
  use:
    needs: analyze
    steps:
      - run: echo "${{ needs.analyze.outputs.result }}"

Conditional execution:

if: contains(github.event.comment.body, '/command')

Resources

• GitHub Actions Documentation
• GitHub REST API
• Anthropic API Documentation
• CODEOWNERS Syntax

Outputs & Deliverables

• Primary Output: GitHub Actions workflows (.github/workflows/*.yml)
• Secondary Output: Automation scripts, CODEOWNERS configuration, documentation
• Success Criteria: Workflows execute successfully in dry-run, automation performs expected actions
• Quality Gate: verification-before-completion evidence (CI logs), guardian review for security-sensitive automations

Constraints

Technical Constraints:

• Do not run production-changing workflows without explicit approval and rollback plan
• Respect GitHub API rate limits (5000 requests/hour for authenticated requests)
• Workflows must complete within timeout limits (default 360 min, recommend <60 min)

Scope Constraints:

• This skill produces GitHub Actions workflows and automation configurations
• Infrastructure provisioning (Terraform, cloud resources) should use ops-manager skill
• Database migrations and schema changes should involve appropriate data skills

Governance Constraints:

• Workflows must respect CODEOWNERS and require appropriate approvals
• Security-sensitive changes require manual review
• Production deployments should use GitHub Environments with protection rules

Common Pitfalls

• Secrets in Logs: Accidentally exposing secrets via echo or error messages → Always use GitHub Secrets and mask outputs
• No Dry-Run Testing: Running automation against production immediately → Test in staging branch/environment first
• Missing Rollback: Automating destructive operations without recovery → Document and practice rollback before deploying
• Flaky Workflows: Race conditions or network timeouts → Add retries and appropriate timeouts to all steps
• Over-Permissioning: Giving workflows more access than needed → Use minimal scopes (read-only unless write required)
• Silent Failures: Workflows fail without notifications → Always configure failure alerts (Slack, email, etc.)
• Ignoring Rate Limits: Hitting GitHub API limits causing automation failures → Implement exponential backoff and caching

Integration Points

Phase	Input From	Output To	Context
Workflow Design	Requirements	GitHub Actions YML	Design automation workflows
PR Automation	Code changes	guardian	Trigger reviews on PRs
Issue Triage	GitHub issues	Labels & assignments	Auto-categorize issues
CI/CD	Test results	Deployment	Run tests, validate, deploy
Verification	Workflow logs	verification-before-completion	Evidence of automation success