DevOps AI Workflows: A Collection of Intelligent Assistant Workflows for DevOps Engineers

Project Name: DevOps AI Workflows (devops-ai-workflows) Core Positioning: A collection of AI agent workflows for DevOps/SRE, encapsulating over 20 common operation and maintenance scenarios (Kubernetes debugging, AWS auditing, Terraform review, CI/CD troubleshooting, etc.). Supported Tools: Mainstream AI programming tools like Claude Code, Cursor, and Windsurf. Source Info: Maintained by 23seriy, published on GitHub (link: https://github.com/23seriy/devops-ai-workflows) on June 3, 2026. Core Value: Transform domain experts' experience into structured workflows, improving operation and maintenance efficiency and consistency.

In daily work, DevOps/SRE often face highly repetitive yet not low-complexity tasks: from Kubernetes cluster fault diagnosis to AWS cost optimization, from Terraform change review to CI/CD pipeline debugging. These tasks require deep domain knowledge and demand fast and accurate execution. This project was born to address these pain points—through structured prompts and rule sets, AI assistants can efficiently handle the above tasks.

The project adopts modular design with a clear directory structure:

1. .claude/commands/: Core workflow definitions (over 20 Claude Code slash commands like /k8s-debug, /aws-cost-quickscan), each including task objectives, steps, and output formats.
2. prompts/: General prompts (event classification, code review, etc.) that can be used in any LLM tool.
3. rules/: Security rule sets to ensure AI follows best practices when executing tasks.
4. scripts/: Independent Shell scripts that support direct use in AI-free environments.

Workflows cover multiple scenarios:

• Kubernetes Ecosystem: 7 workflows (k8s-debug cluster diagnosis, k8s-rbac-audit security audit, helm-release-debug, etc.).
• AWS Cloud Services: 4 workflows (aws-account-audit security audit, aws-cost-quickscan cost scan, etc.).
• IaC: terraform-plan-review (change plan review).
• CI/CD: ci-debug (multi-platform fault diagnosis), jenkins-pipeline-review, etc.
• Security and Observability: secrets-leak-scan (secret leakage scan), repo-health (repository health audit), etc.

Tool Integration Methods:

• Claude Code: Clone the project to the working directory; the system automatically recognizes workflows under .claude/commands as slash commands (e.g., /k8s-debug). After triggering, the AI executes according to predefined steps and generates a report.
• Other Tools: Directly copy prompts from the prompts directory, or refer to security rules in the rules directory to adapt to tools like Cursor and Windsurf.

Take Pod repeated restart failure as an example: The traditional process requires manual execution of kubectl describe pod, log checking, resource analysis, etc., which is time-consuming and prone to omissions. Using the /k8s-workload-debug workflow: AI automatically completes:

1. Obtain Pod status and events
2. Check container error logs
3. Analyze resource requests/limits
4. Verify probe configuration
5. Check storage mounting
6. Review network policies
7. Generate a structured report The time taken is reduced from tens of minutes to a few minutes, ensuring the completeness of troubleshooting.

Project Value:

• Junior Engineers: Provides a structured learning path to quickly master operation and maintenance skills.
• Senior Engineers: Automates repetitive tasks, allowing focus on high-value work.
• Teams: Unified workflows ensure consistency in troubleshooting, reducing issues caused by experience differences.
• Open Source Community: MIT license supports community contributions of new workflows, forming a positive cycle.

Summary: This project is an excellent example of AI-assisted operation and maintenance, deeply integrating LLM with DevOps scenarios. Outlook: In the future, it will cover more cloud services, scenarios, and architecture patterns. Suggestion: Teams can start with common scenarios, gradually integrate AI workflows into daily operations, and build an efficient human-machine collaboration model.