Auditing the Code
A working app is not the same as a production-ready one. Passing a browser test only proves the happy path looks right; it says nothing about what's underneath. So before we deploy, we add a review pass: a fresh session where the agent kicks off multiple sub-agents to audit different parts of the code base, collects what they find, fixes the issues itself, then re-tests the app end to end.
What you'll walk away with:
- Why passing a browser test doesn't mean the code underneath is secure or robust
- Starting a clean session, or clearing the conversation, before the review begins
- Writing a single prompt that triggers a full audit and peer review
- Splitting the review across parallel sub-agents, each covering a different aspect
- Collecting the feedback and having the agent address the issues itself
- Re-testing with the agent's testing tools to confirm nothing broke along the way
- Examples of the review agents that get kicked off, such as a security audit and a Next.js 16 correctness review
- Where this pass belongs in the workflow: right before you push to production