What they did
Rapls maintains a WordPress caching plugin with free and Pro versions. A Pro setting that delays jQuery and inline scripts until the first visitor interaction had lifted PageSpeed mobile scores, so Rapls planned to move it into the free version. The plan was drafted in a chat with Claude and handed to Claude Code as a document. A section at the top told Claude Code to first answer five questions about the current state and stop: how the setting is implemented and named, what Pro excludes by default, how presets treat it, how to avoid double application with both versions installed, and which files would change.
How it went
Claude Code stopped and reported before writing code. Its first finding was that the setting had been in the free version since 1.10.38; the free version only scored lower because its Auto preset did not turn the setting on. Rapls rechecked the original measurements, saw they had compared free-only against Pro with the setting on, and cancelled the move. The new task was to have the free Auto preset enable it, only for tested themes.
Worth knowing
None of the five questions asked whether the work was needed; questions that can only be answered by reading the code were enough to surface the wrong premise. Both the author and the chat assistant had carried the mistaken belief into the handoff. The handoff was written in Japanese and quoted in translation.