Rewrite a vague Claude Code request into a repo-aware task with a concrete verification step.
The starting request, "fix the signup bug please," gives Claude Code no evidence path, no likely search area, and no stop rule beyond sounding finished. Outcome -> hotspot -> scope limit -> verification command The common trap is adding more description while still skipping the only thing that matters: what command or observation will prove the work is actually done. Without that, the agent can produce a polished patch that never touches the real failure mode. Step 1 Name the symptom and the narrow surface: "Investigate why email signup fails after OTP verification in the auth flow." This gives the…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in