Agentic Engineering: The Evolution of Development Practices
Development teams are already using generative AI in their day-to-day work, but typically as an individual exploration: each developer writes their own prompts, without shared standards or true integration into the development workflow.
This course is designed to take the next step: turning that exploration into a standardized team practice that delivers consistent results.
Through a hands-on approach, we will cover everything from the fundamentals of LLMs and agents to the advanced agentic design patterns we use in our own development teams today.
Course information
Format
Online
Duration
10 hours of theory and hands-on practice, split into 3 modules plus optional content
Private student group, tailored to your company
Objectives
- Understand how LLMs and coding agents work.
- Master prompt engineering and context engineering techniques to achieve consistent results.
- Connect the agent to the workspace using tools and MCP, and turn team practices into reusable skills and rules.
- Structure and control complex tasks with agents, incorporating sub-agents with specific roles and independent context.
- Use tests, quality gates, and hooks as mechanisms for verification and aligning the agent's work.
- Incorporate AI-assisted code review to detect vulnerabilities, bugs, and code smells.
- Establish team practices and agreements: what to delegate to the agent, what not to, and how to onboard to the agentic workflow.
- Work securely: permissions, sandboxing, secret management, and preventing destructive operations.
What You'll Learn
- A shared AI framework for the entire team, not just individual skills.
- Criteria for selecting models and tools based on the type of task.
- Concrete verification mechanisms — tests, quality gates, hooks, and AI-assisted code review — to work with agents without compromising on quality.
- Real hands-on practice: each module is practiced on concrete development cases.
Content
Foundations
- LLMs and Agents: model types, capabilities, and selection criteria. Consultative vs. agentic use.
- Prompt Engineering: fundamental techniques and common anti-patterns.
- Context Engineering: context architecture —AGENTS.md and project configurations— and context window mechanics.
Tools and guidelines
- Tools and MCP: agent capabilities and the protocol for connecting them with external systems.
- Skills and Rules: team practices converted into reusable procedures and conventions.
- Security: permissions, sandboxing, secret management, and preventing destructive operations.
- Sub-agents: specific roles, parallel execution, and independent context.
- Team Practices: shared conventions and onboarding to the team's agentic workflow.
AI Workflows
- Workflows: how to structure and control complex tasks with agents.
- The Test as a Contract: agentic TDD, automated testing to align the agent and verify code.
- Quality Gates and Hooks: automated checks to verify and guide the agent's work.
- Agentic Code Review: assisted review to detect vulnerabilities, bugs, and code smells.
- Introduction to Spec-Driven Development (SDD): specifying before building to align intent, code, and tests.
Optional Content (incorporated based on group interests)
- Rapid Functional Prototyping: using generative tools to iterate visual and interactive interfaces (v0, Claude Artifacts, Claude Design, and similar).
- Popular Plugins and Skills: extending agent capabilities through community extensions (Superpowers, Caveman, Skill Creator, among others).
Who Is This Course For?
Software developers and development teams who have already experimented with generative AI tools (ChatGPT, Claude, Copilot) and want to evolve toward a systematic, agentic, team-wide approach.
It also delivers key value to tech leads and architects who need to define adoption standards and criteria for their teams.
It also delivers significant value to tech leads and architects responsible for defining standards and adoption criteria for their teams.
