In this Developing in Agentic AI Systems course, you'll learn:
Foundations of Agentic AI in GitHub
- Define agentic AI in the SDLC and distinguish agents from assistants
- Explain the agent lifecycle: plan, act, evaluate
- Describe GitHub as the system of record and control plane
- Identify responsibilities, risks, anti-patterns, and traceability needs
- Apply the contributor model to agent-generated work
Designing Agent Architecture and SDLC Integration
- Map agent responsibilities to the SDLC
- Define inputs, outputs, and success criteria
- Separate planning, reasoning, and execution
- Implement PR governance with templates, checks, CODEOWNERS, rules, and environment gates
- Build reliable workflows: outputs, contexts, triggers, and cross-job handoffs
- Control and operate agents: observability, tools, MCP, secrets, hooks, and reliability
Tooling, MCP, and Agent Execution Environments
- Understand how agents interact with GitHub APIs and workflows
- Configure Model Context Protocol (MCP) servers, registries, and allow lists
- Define execution context and boundaries (repository, branch, workflow scope)
- Identify agent execution limits and protections
- Enable autonomous actions such as branch and pull request creation
Memory, State, and Evaluation
- Implement agent memory strategies: short-term, long-term, and external memory
- Persist agent state and manage context drift
- Ensure continuity of agent memory and state across tools and environments
- Define evaluation signals and enforce quality gates
- Analyze agent failures and improve behavior
Multi-Agent Systems and Orchestration
- Define multi-agent responsibilities in the SDLC
- Orchestrate agents using GitHub workflows
- Isolate execution: branches, workflows, permissions, and concurrency
- Detect and resolve conflicts using GitHub-native arbitration
- Make the system observable: attribution, evidence, and handoffs
- Operate reliably at scale: diagnose failures and recover safely
Governance, Guardrails, and Operations
- Define risk-based autonomy and action boundaries
- Enforce governance with GitHub controls (rulesets, checks, CODEOWNERS, environments)
- Design human-in-the-loop workflows for high-risk actions
- Control agent capabilities using least-privilege permissions
- Make actions observable, traceable, and auditable
- Maintain governance and operational reliability over time Â
