Prompt Improver Meta-Prompt

Improve the prompt below. It currently produces: [WHAT'S WRONG: e.g. "generic output", "wrong format", "ignores half my instructions", "inconsistent between runs"]

It will be used with: [MODEL/TOOL, e.g. "Claude", "GPT", "a chat UI", "an API call"] for [ONE-TIME USE / REPEATED USE ON MANY INPUTS]

The prompt:
"""
[PASTE YOUR PROMPT]
"""

Rebuild it in three passes:

1. **Diagnose**: list the specific defects, each in one line: vague task verb, missing context or audience, contradictory instructions, no output format, buried key info, persona doing nothing, redundant instructions the model already follows, missing example where format matters, no permission to say "I don't know" on a factual task. Quote the offending line for each defect found.

2. **Rewrite**: the improved prompt, applying only what the defects call for:
   - Open with a direct action verb and the specific deliverable.
   - State who the output is for and why it's needed (context changes output more than adjectives do).
   - Specify the exact output format; add one small example if format matters and none exists.
   - Convert every "don't do X" into "do Y" where possible.
   - Merge or delete contradictory and redundant lines: contradictions silently degrade modern models.
   - Keep it as short as it can be while unambiguous. Do not add structure the task doesn't need.

3. **Explain**: for each change: what you changed → which defect it fixes. Then flag any remaining weakness the prompt owner must fix themselves (usually: missing concrete inputs only they have).

If the prompt will be reused on many inputs, also add: the 3 test inputs I should run first to compare old vs. new, including one edge case likely to break it.