Direct with context.
Turn ambiguous requests into goals, constraints, acceptance criteria, and verifiable work.
Developer masterclass · Two hours
Learn to collaborate with an agentic coding tool without surrendering engineering judgment. The session combines mental models, live repository work, and a repeatable path from request to reviewed diff.
01 / What you will learn
Turn ambiguous requests into goals, constraints, acceptance criteria, and verifiable work.
Make informed permission decisions and recognize risks around commands, credentials, and untrusted content.
Use tests, Git, and independent review to decide whether a result is actually ready.
02 / Curriculum
Designed for developers who know their terminal and Git but are new to agentic coding workflows.
Understand the agentic loop, environment boundaries, installation, and interface.
Write bounded instructions and provide the evidence Claude needs.
Control file, command, network, and external-tool access deliberately.
Use CLAUDE.md, settings, skills, and hooks for durable behavior.
Move from investigation through plan, implementation, tests, and diff review.
Choose appropriately between skills, hooks, subagents, MCP, and plugins.
Use Claude Code to understand, debug, and extend a small agent project.
Recover from drift, troubleshoot failures, and apply an everyday quality checklist.
03 / Format
The class uses a private Google Agent Development Kit project to demonstrate repository orientation, a test-backed bug fix, and a bounded feature implementation.
Prerequisites: basic terminal, Git, and software-development familiarity. No previous Claude Code experience is required.
Last reviewed September 2026
This independent educational material is not affiliated with or endorsed by Anthropic or Google.
Ready to teach the workflow?
Run the masterclass as a focused two-hour session for engineering teams, technical communities, or internal developer programs.
Discuss a session