Resources
Playbooks, guides, and best practices for AI-native E2E testing.
CI/CD for Agent-Written Code
When coding agents open most of your pull requests, unit tests passing is no longer proof the change is safe to merge. This guide covers the pipeline stage AI-heavy PR flows are missing: a PR-time behavioral gate that verifies each change in a real browser before it merges, with GitHub Actions setup notes.
Context Engineering for Coding Agents
Context engineering is the practice of curating exactly what a coding agent sees at the moment it acts: instructions, repository knowledge, tools, tests, and runtime feedback. This guide breaks down the kinds of context, the principles for managing them, and why a checkable definition of correct is the highest-value context you can supply.
Continuous Verification for AI-Generated Code
Continuous integration proved the code compiles. Continuous testing proved the existing tests pass. Continuous verification proves that each change actually behaves the way it was meant to, on every change, at the speed AI coding agents now produce them.
How to Test Code Written by Cursor
Cursor writes and edits UI code fast, but it rarely confirms the result works in a browser. This guide shows how to give Cursor a real browser to verify its own changes and author maintained YAML end-to-end tests that live in your repo.
How to Test Code Written by Antigravity CLI (and Gemini CLI Before It)
Google retired Gemini CLI on 18 June 2026 and replaced it with Antigravity CLI. The verification loop is the same either way: run the change in a real browser, save the check as a durable E2E test, and gate every merge in CI.
How to Verify AI-Generated Code: A Practical Checklist
Verifying AI-generated code means proving the change does what it was asked to do, not just that it reads well and passes unit tests. This checklist walks the full sequence: static review, unit tests, integration and contract checks, behavioral verification in a real browser, and a regression test that stays green. The behavioral step is the one most teams skip, and it is where the failures that reach users hide.