AI Free course · Certificate included
AI-Assisted Software Development: Coding Agents in Practice
Ship reviewed, tested code with Claude Code, Codex, Copilot and Cursor: context engineering, guardrails, CI and governance
- Intermediate
- 7 h 5 min
- 17 lessons in 6 modules
- 2 h 30 min of video lectures
- Updated Sep 2026
About this course
Coding agents can now read your repository, edit code, run tests and open pull requests on their own. Used well, they compress days of work into hours; used carelessly, they ship subtle bugs, leak secrets and flood reviewers. This hands-on course teaches the engineering discipline that makes agents safe and genuinely productive. You will map the 2026 landscape (Claude Code, OpenAI Codex, GitHub Copilot agent mode and cloud agent, Cursor, Devin Desktop, Gemini Code Assist and Antigravity), master context engineering with AGENTS.md and CLAUDE.md, and run plan-implement-verify loops with tests as guardrails. You will use AI for code review, large refactors and debugging, wire agents into GitHub Actions, defend against secret leaks and prompt injection from repositories, set team governance and measure productivity honestly. The capstone ships a real feature end to end with an agent, tests included.
Tools you’ll use
- Claude Code
- OpenAI Codex
- GitHub Copilot
- Cursor
- Devin Desktop
- Gemini Code Assist
- Google Antigravity
- GitHub Actions
- AGENTS.md
- Model Context Protocol
- pytest
- Vitest
- gitleaks
Skills
- AI-assisted development
- Coding agents
- Context engineering
- Test-driven development
- AI code review
- DevSecOps
- Engineering governance
What you’ll be able to do
- Choose the right coding tool and interaction mode (pair, delegate, pipeline) for a given task
- Write AGENTS.md/CLAUDE.md instruction files and task specs that agents can execute reliably
- Run plan-implement-verify loops with tests, hooks and small diffs as guardrails
- Use agents for code review, debugging and large refactors or migrations without losing control
- Integrate coding agents into GitHub Actions with least-privilege permissions
- Defend against secret leaks, risky dependencies and prompt injection hidden in repositories
- Set team governance and measure AI-assisted productivity with honest metrics
Curriculum
Syllabus
- Modules
- 6
- Lessons
- 17
- Reading time
- 4 h
- Assessment questions
- 30
Map the 2026 coding assistant and agent landscape, understand the agent loop and context limits, and choose and set up tools through a fair trial.
- From autocomplete to agents: the 2026 landscapeVideo lecture, 9′12 min
- How coding agents work: loops, context and permissionsVideo lecture, 9′13 min
- Choosing and setting up tools with a fair trialVideo lecture, 9′12 min
Give agents the right context: instruction files (AGENTS.md, CLAUDE.md), repo maps, skills, MCP and subagents, and task specs that agents can actually finish.
- Instruction files: AGENTS.md, CLAUDE.md and friendsVideo lecture, 9′13 min
- Repo maps, skills, MCP and the context budgetVideo lecture, 9′13 min
- Writing tasks agents can finishVideo lecture, 9′12 min
Run coding agents through explicit explore-plan-implement-verify loops, use tests as the core guardrail, and debug with a scientific protocol.
- The explore-plan-implement-verify loopVideo lecture, 9′14 min
- Tests as guardrails: TDD with agentsVideo lecture, 9′14 min
- Debugging with agents: a scientific protocolVideo lecture, 8′13 min
Use AI for code review without drowning in noise, run large refactors and migrations safely, and integrate agents into GitHub Actions with least privilege.
- Code review with AI, and reviewing AI codeVideo lecture, 9′13 min
- Refactoring and migrations at scaleVideo lecture, 9′14 min
- Integrating agents into CI with GitHub ActionsVideo lecture, 9′14 min
Defend against secret leaks, risky dependencies and prompt injection; roll out agents with a clear policy; and measure productivity with balanced, honest metrics.
- Security: secrets, dependencies and prompt injectionVideo lecture, 9′15 min
- Team adoption and governanceVideo lecture, 8′13 min
- Measuring AI-assisted productivity honestlyVideo lecture, 9′13 min
Know when vibe coding is fine and when it isn't, harden prototypes deliberately, and ship a real feature end to end with an agent, tests and review.
- Vibe coding vs engineering disciplineVideo lecture, 9′12 min
- Capstone: ship a feature end to end with an agentVideo lecture, 8′20 min
- Final assessment
Your certificate
Finish with a credential anyone can check
Earn the Certified AI-Assisted Software Engineer badge: The holder can deliver production software with coding agents safely: choosing tools and modes, engineering context with instruction files and task specs, running plan-implement-verify loops with tests as guardrails, using AI for review, debugging and migrations, integrating agents into CI with least privilege, defending against secret leaks and prompt injection, and measuring productivity honestly.
Completed all lessons and scored at least 80% on the final assessment.
- A public verification page
- A PDF certificate to download
- An Open Badge you can share
- One click to your LinkedIn profile
Final assessment
- 30questions drawn from a larger pool
- 45 mintime limit
- 80%pass mark
- 3attempts per 24 hours
Start learning today. It’s free.
Every lesson is free to read. A free account saves your progress, unlocks the final assessment and issues your certificate.