Multi-Agent Novel Writing System

What it solves

When generating long-form fiction directly, models often drift in three ways: the premise never settles, characters and world rules become inconsistent, and chapter pacing misses the target platform. This skill treats novel writing as a reviewable workflow rather than a one-shot generation task: clarify readers and platform, split the project into premise, world, characters, structure, and style, then draft and revise chapters in controlled loops.

How it works

novel-writer-team simulates an eight-role writing team covering creative direction, worldbuilding, character design, outlining, style, drafting, editing, and project control. The workflow first collects inspiration, platform, length, and preferences, then serializes premise, world, and character setup. Outline and style can be produced in parallel before a project review. Chapter drafting uses a write-review-revise cycle; the editor checks plot logic, character consistency, pacing, foreshadowing, platform fit, hooks, and prose quality. Each chapter is limited to three review rounds, with the project lead stepping in if issues remain. Project references such as references/agents.md, platforms.md, workflows.md, templates.md, and example-project.md are loaded on demand to keep context focused.

Where it fits

It works best for structured novels or serialized fiction where the target platform, chapter length, and style constraints matter. If the goal is a short, free-form essay or highly experimental prose, the confirmation checkpoints may be too heavy. The output still depends on the quality of the user’s premise, the specificity of platform rules, and whether the user can review key milestones.