bookmorph-magic（书形魔法）

BookMorph Magic

Use this skill as a publishable template for book-to-content orchestration. It is not a private, fixed workflow.

What this template does

• Takes one book or one episode-style request
• Produces a package of multi-format content
• Bundles the final outputs into a stable episode directory with a manifest

Typical outputs:

• one video
• one audio file
• three cover images (3:4, 4:3, 1:1)
• one manifest.json

Important constraints

• This is a template, not a turnkey SaaS workflow.
• To run end to end, you must provide 3 adapters:

- book selector - longform content generator - cover generator

• If you already have existing skills or scripts for those adapters, wire them to the contracts in references/integration-template.md.

Inputs

• A book file path, or
• A request containing one episode identifier such as ep38, EP38, episode 38, or 第38期

Output contract

• Output root is configurable
• Episode directory prefix is configurable; default is episode-
• Stable output filenames:

- epNN-video.<ext> - epNN-audio.<ext> - epNN-cover-3x4.<ext> - epNN-cover-4x3.<ext> - epNN-cover-1x1.<ext> - manifest.json

Workflow

1. Parse the episode identifier with scripts/episode_bundle.py parse-episode.
2. Create or reset the target episode directory with prepare-dest.
3. Run your book selector adapter until it chooses exactly one source book.
4. Run your longform content generator adapter to produce a video and an audio output.
5. Run your cover generator adapter to produce 3:4, 4:3, and 1:1 images.
6. Run bundle to copy the final artifacts into the episode directory and write manifest.json.

Adapter requirements

Book selector

Must return enough information to identify one selected source:

• book title
• author
• local source file path
• selection round or attempt count
• optional selector checkpoint path

Longform content generator

Must return:

• video file path
• audio file path
• attempt count
• optional generator checkpoint path

Cover generator

Must return:

• image path for 3:4
• image path for 4:3
• image path for 1:1
• attempt count
• optional prompt archive path
• optional generator checkpoint path

Failure rules

• If the episode number cannot be parsed, fail immediately.
• If no source book can be selected, stop and write a failed manifest.
• If longform media is incomplete, stop and write a failed or partial manifest.
• If any required cover image is missing, stop and write a failed or partial manifest.

Commands

python3 {baseDir}/scripts/episode_bundle.py parse-episode --text "制作ep38内容" --output-root "/abs/output/root"
python3 {baseDir}/scripts/episode_bundle.py prepare-dest --episode 38 --clear-existing --output-root "/abs/output/root"
python3 {baseDir}/scripts/episode_bundle.py bundle --episode 38 --video "/abs/video.mp4" --audio "/abs/audio.m4a" --image-3x4 "/abs/3x4.png" --image-4x3 "/abs/4x3.png" --image-1x1 "/abs/1x1.png" --book-title "Example Book" --author "Example Author" --book-path "/abs/book.epub" --selector-adapter "my-book-selector" --selector-reference "round-1" --selection-round 1 --longform-attempts 1 --cover-attempts 1 --prompt-archive "/abs/prompts.md" --output-root "/abs/output/root"
python3 {baseDir}/scripts/episode_bundle.py bundle --episode 38 --selection-round 3 --longform-attempts 0 --cover-attempts 0 --status failed --failure "selector=no valid source book" --output-root "/abs/output/root"

References

• Adapter contracts: references/integration-template.md
• Manifest fields: references/manifest-schema.md