The speaker is using Claude, a coding agent, to audit their React code for best practices in a desktop application they are building. They admit they are not a React expert and are likely not following best practices. They showcase the "React Best Practices" skill among many custom skills they've developed for code auditing and project setup. The speaker specifically invokes the React Best Practices skill to audit a "sidebar" component, suspecting performance issues due to improper React usage like missing useMemo or incorrect dependencies.
The audit, using the Vercel React Best Practices skill, took over five minutes and revealed several violations in the sidebar code. These include "No React compiler detected," issues with process hotkey assignments causing re-renders, subscribing to entire objects instead of specific properties, lack of memoization, and misleading comparators. The speaker recognizes the value of the skill in identifying domain-specific issues the LLM-written code has.
The next step the speaker outlines is to put Claude into "plan mode" to create a detailed plan to address the identified issues. This involves reviewing the plan, ensuring it makes sense, and then having Claude execute it, including adding tests, running tests, and pushing to CI. The video concludes by stating the next topic will be discussing where the skills are located on the machine.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.