Skip to content

Latest commit

 

History

5 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Gemini Workflows: Agentic Software Development

A Multi-LLM Workflow Suite for the AntiGravity IDE.

This toolkit provides a set of specialized, "best-in-class" agentic workflows designed to leverage the unique strengths of the three leading AI models: Claude Opus 4.5, GPT-5.2, and Gemini 3 Pro.

🧠 The "Big Three" Strategy

Instead of using a single model for all tasks, these workflows assign the right "brain" to the right job:

Role Agent Model Why?
Lead Engineer /agent_backend_expert Claude Opus 4.5 80.9% SWE-bench score. Best for deep coding, refactoring, and logical correctness.
Frontend Expert /agent_frontend_expert Claude Opus 4.5 Handles complex React state, hooks, and large-scale refactors with high precision.
Project Manager /create_jira_ticket Claude 3.5 Sonnet Fast, concise, and excellent at natural language task breakdown.
Infra Architect /agent_infrastructure_expert GPT-5.2 (Thinking) "Thinking" Mode excels at orchestration, distributed systems logic, and avoiding race conditions.
Chief Auditor /agent_audit_expert Gemini 3 Pro 1M+ Token Context. Can read entire repositories and logs to find subtle security/compliance issues.

📂 Workflow Manifest

  • agent_backend_expert.md: Backend implementations (Python/Go/Node).
  • agent_frontend_expert.md: Frontend implementations (TSX/React).
  • agent_infrastructure_expert.md: Cloud & DevOps (Terraform/K8s).
  • agent_audit_expert.md: Full-repo security & compliance audits.
  • audit_backend_functional.md: Backend functional goal verification.
  • audit_backend_security.md: Backend security & HIPAA checks.
  • audit_backend_technical.md: Backend code quality & metrics.
  • audit_frontend_functional.md: Frontend UX flow verification.
  • audit_frontend_security.md: Frontend XSS & dependency checks.
  • audit_frontend_technical.md: Frontend performance & compatibility.
  • skill_code_review.md: Automated code quality checks (Opus 4.5).
  • skill_test_gen.md: Automated test case generation (Opus 4.5).
  • create_jira_ticket.md: Jira ticket creation helper (Sonnet 3.5).
  • backend_deploy.md: Production deployment orchestration pipeline.

🚀 Usage

Import these workflows into your .agent/workflows directory to instantly upgrade your AntiGravity IDE agent with specialized personas.

"The future of coding isn't one model—it's a team."

💿 Interactive Installation

We provide scripted installers for a seamless setup experience. These scripts will:

  1. Copy the workflows to your preferred directory (default: .agent/workflows).
  2. Check for the existence of your .env file.
  3. Verify that required API keys (ANTHROPIC_API_KEY, OPENAI_API_KEY, GOOGLE_API_KEY) are present.
  4. Provide color-coded status updates.

Mac / Linux / Unix

Run the bash script:

chmod +x install.sh
./install.sh

Windows

Double-click install.bat or run it from Command Prompt:

install.bat

🔗 Chaining Strategy: End-to-End Flow

For a complete feature lifecycle, chain the agents in this order:

1. Implementation (Opus 4.5)

  • @agent_backend_expert "Design a User API"
  • @agent_frontend_expert "Connect API to Dashboard"

2. Audit & Verification (Gemini 3 Pro)

  • @agent_audit_expert "Audit the auth module for security"
  • @skill_test_gen "Generate tests for user_service.py"

3. Deployment (GPT-5.2)

  • @backend_deploy "Deploy the new user service"
  • @create_jira_ticket "Task: Verify user flow in prod"

✍️ Author & Credits

Dasa Elayavilli Ponnappan UC Berkeley MIDS '23 | AI-ML Enthusiast | Fractional CTO | Serial Entrepreneur with Successful Exit

📧 Contact: dasa@manizh.com, epdasa@berkeley.edu

About

Agentic AI workflows that enable supercharged AI Native development team providing persistent memory to your Agent. Crafted for Google AntiGravity Agent..

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages