scene-composition

To partition a branching story tree into scenes and acts, provide the full node set JSON. The skill groups beats by shared pressure or motif, returning a JSON object with labeled scenes, motifs, and act numbers.

Scene clustering in narrative story trees is the process of grouping 6–40 beats into 2–7 cohesive scenes. It uses deterministic partitioning based on shared setting, pressure, or motif to reveal latent act structures.

Yes, you can label story motifs and acts from an existing branching JSON structure. The skill processes the provided beats and assigns a story-specific title, motif, and act number to each partitioned scene without modifying original beats.

Scene clustering works with branching story trees containing 6–40 beats. It groups these nodes into 2–7 scenes, making it suitable for small to medium interactive narratives rather than single linear scripts.

You extract scene structure from a story tree without altering original nodes by processing the JSON input read-only. The skill returns exactly one new JSON object containing the scenes array, ensuring the original beats remain completely unmodified.

Limitations of deterministic narrative scene clustering include strict node count requirements; the story tree must have 6–40 beats to output 2–7 scenes. Trees outside this range lack sufficient pressure grouping data for valid partitioning.