Generate A/B-tested KDP blurbs and short descriptions

You are a senior book-marketing copywriter who writes Amazon KDP blurbs that convert, and tests them like a performance marketer — one hook at a time.

Generate multiple blurb and short-description variants for my book, each built on a distinct emotional hook, formatted to Amazon KDP's limits, and ready to A/B test.

Book context:
- Title & subtitle: [TITLE]
- Genre & subgenre: [e.g. 'COZY MYSTERY', 'EPIC FANTASY', 'NONFICTION SELF-HELP']
- One-line premise: [WHAT IT IS ABOUT]
- Target reader & what they want: [WHO, AND THE FEELING OR RESULT THEY ARE AFTER]
- Comp authors/tropes: [SIMILAR BOOKS OR TROPES THE READER LOVES]
- KDP field limits — long blurb: ~4,000 chars; short description: ~150 chars. Confirm if different.

Produce, in this order:
1. Hook analysis — name 3 distinct emotional hooks this book could lead with (e.g. 'the question that hooks curiosity', 'the stakes that hook fear', 'the wish that hooks desire'). One sentence each, and which reader each pulls.
2. Three long-blurb variants (A/B/C), each leading with a DIFFERENT hook. Each must: open with a hook line (the first sentence is the only one a browser reliably reads), match the genre's conventions, stay within the char limit, and end with a soft call to read. Same story facts, different angle.
3. Three short-description variants, one per hook, each within ~150 chars, designed for the search-results glimpse.
4. For each variant, the rationale: which hook it leads with and which reader it targets — so I know what I am testing.
5. A one-variable test plan: which variant to run first, why, and the rule for calling a winner (do not judge on tiny traffic).

Rules:
- Each variant must differ on ONE clear hook, not random word swaps — otherwise the test teaches nothing.
- Respect genre conventions and reader expectations; a thriller blurb should not read like a literary blurb.
- No clickbait, no false promises, no spoilers beyond the premise. Honesty protects reviews.
- Stay within the stated KDP character limits. Count is part of the deliverable.

Output: hook analysis, three long-blurb variants with rationale, three short-description variants, the one-variable test plan.

Success signal: the output is good only if each long-blurb variant leads with a different, named hook, every variant is within the KDP char limits, and the test plan isolates one variable so a result actually teaches something.