What changed in the rules, and why.

Promoted the rule book to canonical source

The producer rules are no longer just a review prototype. The pipeline now reads and syncs from bookmaking_rules.html.

• Added rules_registry.py to validate the embedded rule block and regenerate derived files.
• Generated rules_data.js for the review UI and generated_prompt_snippets.md for prompt surfaces.
• Wired story generation, manuscript tasks, critique tasks, image prompt rubrics, and book configuration editorial defaults back to the rule registry.
• Added tests that fail if generated rule surfaces drift from the canonical rule book.

Imported Rabat style insights into active rules

The final Rabat English PDF is now the benchmark for future manuscript density and page-design variety.

• Changed the spread word target to 40-60 words, with 70 as a soft ceiling and 75+ as an editorial review flag.
• Added spread text roles: full text, short caption, sound-effect moment, and image-led beat.
• Added prompt guidance for each spread to name its visual job, physical joke, and freshness against repeated city motifs.
• Clarified that Nunito remains the default story-text font, while sound effects and emphasis may receive expressive display treatment.

Added freshness, comedy, and local-animal rules

Story generation now has explicit guidance to compare against prior approved stories before choosing tropes, guest animals, and comic engines.

• Added a freshness rule: repeat Toto's lovable mannerisms, not the same plot machinery.
• Added a comedy rule for whimsy, laugh-out-loud physical humor, and Toto's sweet naivety.
• Added a local-animal rule: prefer city/country-associated animals and avoid repeating the same animal in multiple stories for that country.
• Extended the variation register to remember guest species and guest-character comic engines.

Moved typography and dimensions into book configuration

The rules now treat Nunito and print dimensions as inherited book settings rather than hardcoded rule text.

• Removed the extra Copy Rule Block button; Copy Prompt is the single producer action.
• Changed page-text guidance to use the configured page-text font, defaulting to Nunito.
• Removed the hardcoded 2:1 cover rule and marked cover sizing as configuration-driven.
• Confirmed the 8 x 10 Amazon KDP default: 1536 x 1024 px GPT Image spreads, 2400 x 3000 px trim page, 2438 x 3075 px bleed page, and 4875 x 3075 px two-page spread source at 300 PPI.

Shifted the live rules toward producer-facing creative decisions

The main review UI no longer foregrounds technical workflow language. It now starts with story setup and adds rules for image-prompt intent, text placement, active Toto poses, and avoiding repeated city motifs.

• Replaced the technical queue-first rule with a reviewable story idea rule.
• Added a story-choice checklist so generated stories expose producer decisions before prose.
• Added Lisbon-derived guidance that motifs such as trams should not dominate too many spreads without a story reason.
• Added a rule that Toto should be visibly active in every spread.

Moved the rules prototype into one folder

The review UI, editable rules source, history, and preview now live together in docs/rules/.

• index.html is the review and editing surface.
• bookmaking_rules.html is the plain editable source.
• history.html keeps retired decisions out of the live rule list.

Removed resolved conflicts from the active rules

The live source now describes current behavior only. Older or rejected behavior moved here so the day-to-day rules stay short.

Added evidence weighting

Durable rules should lean most on Lisbon production work, then Rabat reusable lessons. New Delhi remains useful as demo data, but should not drive permanent rules by itself.