What problem does it solve? Writing a new chapter of a serialized web novel requires the author to juggle worldbuilding documents, character cards, foreshadowing trackers, error logs, and previous chapters. Manually assembling this context is slow and error-prone, and feeding everything raw to an LLM blows the context budget. This Skill automates the collection, prioritization, and compression of all relevant materials into a single structured context block. ## Core Features & Use Cases - Structured context assembly: Reads worldbuilding docs, character cards, foreshadowing tables, storyline status, outlines, and the last 1-5 chapters, with explicit handling for first-chapter (N=1) creation mode. - Weighted error recall: Applies a decay formula (weight × 0.7^chapter distance) to past correction records so only recent, high-impact mistakes surface in the context. - Foreshadowing prioritization: Sorts pending plot threads by overdue status, urgency tier, and remaining chapters so critical setups are never dropped. - Token budget compression: Enforces a 2000-token budget with a strict priority ladder that never trims character state, plus a reader-retention (drop-off) prediction for the target chapter. - Use Case: Before drafting chapter 42 of a fantasy serial, run this Skill to get a compact brief showing each character's current state, the two overdue foreshadowing threads due this chapter, the outline excerpt, the previous chapter's cliffhanger, and the top weighted correction feedback. ## Quick Start Prepare the writing context for chapter 42 of my novel using the novel-prep skill.
