Naming, not cover design

Never one winning title. Always the safe pick, the bold swing, and the risk beside each one.

Two coupled jobs: naming a book before it exists, and retitling one that's already out and not finding readers — plus the jacket copy that runs on the same thinking. Proven four times across a real shelf: two rounds on one book, and an identical retitle workflow on three others.

See the two jobs View on GitHub

Two jobs. Different entry conditions, same discipline underneath.

Which one applies depends entirely on whether the book already has a published title.

No published title yet

First-time development

5 candidates a round, each with a why-it-works/risk pair, against a stated north star and the book's real constraints. A new round runs when the manuscript's actual thesis evolves — not just on request.

  • Safe pick, bold commercial swing, dark horse — never one winner
  • Grounded in a real two-round trail, ten real candidates
  • Feeds the jacket-copy discipline and the shelf check
Already published, not converting

Retitle-for-positioning

The exact _archive-old-title/ convention: old PDFs/PNGs archived, codenames and folder names untouched, only the displayed title text changes.

  • Proven three separate times, no deviation
  • Starts from a diagnosed cause, not a guess
  • Same why-works/risk discipline applied to the retitle framing
The refusal gate this skill runs on

"I won't hand you a single 'winning' title — every candidate ships with what makes it work and what could go wrong, and the recommendation names a safe pick, a bold commercial swing, and a dark horse, not one verdict." The candidate that looks "obviously the strongest" is usually the boldest swing — and the boldest swing is, by construction, the most exposed.

The comparable-title scan — formalized, not assumed

The real trail flagged shelf-collision risk informally, in individual risk notes. This build ran the check for real, live, on the same category one of those notes named.

Real shelf data, fetched 2026-09-22 (Goodreads "fasting" shelf — Amazon's own list returned a 503 this session):

  1. Autophagy: ...to Promote Longevity, Optimal Cellular Renewal, Detox, and Strength for a Happy Life
  2. The Complete Guide to Fasting: Heal Your Body Through Intermittent, Alternate-Day, and Extended Fasting
  3. Fast Like a Girl: ...to Burn Fat, Boost Energy, and Balance Hormones
  4. A Hunger for God
  5. Fast, Feast, Repeat: The Comprehensive Guide to Delay, Don't Deny® Intermittent Fasting
  6. The Obesity Code: Unlocking the Secrets of Weight Loss
  7. Life in the Fasting Lane: ...Simple, Sustainable, and Enjoyable
  8. Fasting Can Save Your Life
Dominant pattern: outcome-promise-in-subtitle — burn fat, boost energy, unlock the secrets. Are You Actually Hungry? — a plainspoken question, no outcome-promise — is a genuine mismatch against this shelf. That confirms, with real data, what Round 1's own risk note predicted informally. The eventual subtitle doesn't chase the pattern either — a deliberate pattern-interrupt, not an accidental one. Naming that difference is exactly what this step exists for.

Jacket copy: dated, versioned, and checked against the book it's actually shipping on

The same "describe it to a stranger in one breath" thinking that produces a title produces the back cover.

A real incident on this shelf

A publisher's comp once carried a different book's back-cover text verbatim — one title's copy, word for word, on another book's proof. It wasn't caught by assuming the comp was correct. It was caught by checking the returning proof character-for-character against the dated copy-of-record file. Every proof gets that same check now, not just the one that failed once.

Titles shipped on this discipline

Proof of shipped retitles and locked titles goes here — added by the operator, not invented.

Setup

  1. Load this folder into a Claude Project, or point a Claude Code session at it — SKILL.md lets it auto-discover and trigger from a natural request.
  2. Say which situation applies: no published title yet, or an already-published book that isn't converting.
  3. Bring the manuscript's real thesis (not a pitch) for first-time development, or a stated reason for a retitle — the specialist refuses to guess at either.
  4. Let it run the comparable-title scan before locking anything, and review the safe/bold/dark-horse recommendation rather than picking on hook alone.
  5. Hand the locked title to your layout and cover skills — this one doesn't typeset or design.

FAQ

Will it just tell me the best title?

No. Every candidate ships with its risk, and the recommendation names three real options, not one verdict — see the refusal gate above.

Does it design the cover?

No. It supplies the title and description text a cover has to carry — cover visual execution is a separate, deferred skill.

Does it write the whole launch/funnel plan?

No — that's Book Launch & Funnel Strategy. This one owns the title/description artifact, though it feeds that larger strategy.

What if I ask for a retitle without saying why it's not working?

It asks what evidence that's based on before proposing anything — a retitle without a diagnosed cause risks solving the wrong problem.