Agentic AI as a Part of Software Development - Learning Roadmap | Nemorize

✅ Foundations of Agentic AI

• ✅ Agent Anatomy & Patterns
  • ⚪ LLM + Tools + Memory + Loop
  • ⚪ ReAct Pattern
  • ⚪ Assistance → Autonomy Arc
• ✅ Context Engineering
  • ⚪ Five Context Layers
  • ⚪ Attention Budgets & Failure Modes
  • ✅ AGENTS.md & CLAUDE.md

✅ Prompt Chaining & Parallel Workflows

• ⚪ Prompt Chaining Patterns
  • ⚪ Sequential & Router Chains
  • ⚪ Structured Outputs as Glue
  • ⚪ Chain Debugging
• ⚪ Fan-Out & Fan-In Workflows
  • ⚪ Map-Reduce & Scatter-Gather
  • ⚪ Partial Failure Handling
  • ⚪ Cost & Latency Math

✅ Protocols, Tools & Skills

• ⚪ MCP & A2A Protocols
  • ⚪ MCP Servers & Gateways
  • ⚪ A2A Agent Cards
• ⚪ Tool Design & Skill Loading
  • ⚪ Tool Design Principles
  • ⚪ SKILL.md & Progressive Loading
  • ⚪ Self-Authored Skills

✅ Building & Operating Production Agents

• ⚪ Telemetry & Observability
  • ⚪ Tracing & Span Attributes
  • ⚪ Key Production Metrics
• ✅ Testing & Evaluation
  • ✅ Testing Pyramid for Agents
  • ✅ LLM-as-Judge & A/B Evals

✅ Agents Across the SDLC & Frontier

• ✅ Security & Architecture
  • ⚪ Prompt Injection & Sandboxing
  • ⚪ Retrofitting & Strangler-Fig
• ⚪ SDLC Integration & Frontier
  • ⚪ Agents Across the SDLC
  • ⚪ Team Shifts & Frontier Practice