Skip to content

An AI skill that turns books into animated, narrated, interactive web experiences

docs(skill): ask for models as one choice after the tier

DozenTwelve07cdba27ef18
diff --git a/.claude/skills/papermorph/references/core/style.md b/.claude/skills/papermorph/references/core/style.md
index c25c861..615a50a 100644
--- a/.claude/skills/papermorph/references/core/style.md
+++ b/.claude/skills/papermorph/references/core/style.md
@@ -12,9 +12,14 @@ Left to guess what the user wants, an agent's guesses diverge from run to run. S
## Models
-Which model does which step is the user's choice, among the models the host running the Skill can start; a user who wants another provider's models makes the book in that provider's host. Show the tier's defaults as one table, one row per step of its team ([core/team.md](team.md)), and ask: keep it (recommended), or change some rows.
+Which model does which step is the user's choice, among the models the host running the Skill can start; a user who wants another provider's models makes the book in that provider's host. How much animation the book has is already settled by the tier the user chose at intake; the models come after it, as one question the user answers with a click:
-The defaults are the splits that made approved books; a cheaper split is the user's to try.
+- **Recommended**: the tier's column of the table below, the splits that made approved books.
+- **Strongest everywhere**: every step on the strong model.
+- **Cheaper**: for a lesson, the picture step on the middle model and the rest as recommended, tried on tier 1 only (below). For a picture book, the script on strong, sound on small and every other role on middle. For a film, the builder (a documentary's one agent, a short's worker per episode) on middle, and the fresh director still strong. The tier-2, picture-book and film splits have not been tried yet; say so in the option.
+- **Custom**: only then show the table, one row per step of the tier's team ([core/team.md](team.md)), each row offering the host's models to pick; at four questions a round, a lesson's seven rows take two rounds.
+
+A film's recommended split already is strongest everywhere, so a film offers three options: strongest everywhere (recommended), cheaper and custom. Lessons and picture books offer all four, recommended first and labelled so.
| Step | Lesson (tier 1 and 2) | Picture book | Documentary | Short |
|---|---|---|---|---|
@@ -27,9 +32,13 @@ The defaults are the splits that made approved books; a cheaper split is the use
| Register | small | small | strong (the coordinator) | strong (the coordinator) |
| Made this way | physics ch14, at about half the cost of one strong agent | *Joke Collection* episodes 5–6 | *Madness and Civilization*, *The Pandemic* | *Madness and Civilization* shorts 2–4, *The Pandemic* shorts 1–6 |
-A middle model on a lesson's picture step made two maths chapters that were acceptable but monotonous: a cheaper choice for tier 1 when the user wants it. No cheaper split of a film has been tried.
+A middle model on a lesson's picture step made two maths chapters that were acceptable but monotonous: a cheaper choice for tier 1 when the user wants it.
+
+**small**, **middle** and **strong** stand for the user's current fast, balanced and most capable models. Each option names the actual models, not only their strength: Claude Code starts a subagent on `opus`, `sonnet`, `haiku` or `fable`, so a lesson's recommended option reads "opus for the picture and rework, sonnet for narration and review, haiku for the rest". Where you cannot tell which models the host can start, ask the user once which is their strong, middle and small model. A host that cannot choose a subagent's model, or has no subagents, runs every step on the conversation's model, so say that to the user and let the table record it. A model the host cannot start is named to the user, not swapped for another.
+
+Every step runs at high reasoning effort unless the user asks for another; effort is not one of the questions.
-**small**, **middle** and **strong** stand for the user's current fast, balanced and most capable models; write the actual names into `models`, keyed by step as in the [lesson](../../assets/templates/style.json) or [picture-book](../../assets/templates/picture-book/style.json) template (a picture book's keys name its roles: `script`, `kit`, `scenes`, `sound`, `director`, `reviewers`; a [documentary](../../assets/templates/documentary/style.json)'s and a [short](../../assets/templates/shorts/style.json)'s: `script`, `pictures`, `beats`, `checks`, `director`). Claude Code starts a subagent on `opus`, `sonnet`, `haiku` or `fable`; a host that cannot choose a subagent's model, or has no subagents, runs every step on the conversation's model, so say that to the user and let the table record it. A model the host cannot start is named to the user, not swapped for another. The user can change a row at any time; later chapters follow the table as it stands.
+Write the actual names into `models`, keyed by step as in the [lesson](../../assets/templates/style.json) or [picture-book](../../assets/templates/picture-book/style.json) template (a picture book's keys name its roles: `script`, `kit`, `scenes`, `sound`, `director`, `reviewers`; a [documentary](../../assets/templates/documentary/style.json)'s and a [short](../../assets/templates/shorts/style.json)'s: `script`, `pictures`, `beats`, `checks`, `director`), and the option the user chose, in its wording, in `words`. The user can change a row at any time; later chapters follow the table as it stands.
## style.json
@@ -38,7 +47,7 @@ Start from the tier's template: [lessons](../../assets/templates/style.json), or
| Block | What it holds | Present in |
|---|---|---|
| `schema`, `tier`, `book` | schema version; `1`, `2`, `"custom"`, `"picture"`, `"documentary"` or `"shorts"`; title, slug, language, readers, pilot chapter | all |
-| `words` | the user's own words about the look, verbatim, with later feedback that changed the bar | all |
+| `words` | the user's own words about the look, verbatim, with later feedback that changed the bar, and the models option they chose | all |
| `look` | palette, what each colour means, text and maths fonts, notation and rounding | all |
| `voice` | TTS engine (`edge`; `kokoro` when Edge is unavailable or local audio is wanted; `eleven`, paid, for a film: [core/sound.md](sound.md#voices)), voice, rate | all |
| `pacing` | minutes per chapter, where checks and practice go | all |