One of the most common prompt engineering mistakes is treating every task the same. Simple tasks get overcomplicated prompts. Complex tasks get underpowered prompts. Both produce worse results than calibrating correctly.
For fast, low-stakes outputs — a subject line alternative, a brief follow-up nudge, a quick status update — keep the prompt short and direct. Adding context and instructions that don't apply to a simple task actually reduces output quality by diluting Claude's focus.
"Write 3 subject line alternatives for this email: [PASTE EMAIL]." That's a complete, appropriate prompt for that task.
For complex, high-stakes outputs — an executive summary, a MEDDIC analysis, a full objection response with proof points — you need more prompt. Add role, context, format, length constraints, and tone requirements. The output quality scales directly with the quality of your setup.
If your prompt is longer than your expected output, it's probably too complex for the task. If your output needs to be longer than your prompt, you probably didn't give Claude enough to work with.