Teacher Luo Novel Prompt Engineer

Problem

Long-form AI writing often fails because prompts mix layers instead of keeping fixed settings, current state, author intent, task, and output constraints separate. This causes the model to treat suggestions as facts, old state as present, and rhetoric as plot.

How It Works

The skill turns writing into an auditable process:
- Five-layer context: fixed facts, current state, creative intent, current task, and output constraints.
- Work modes: ideation, outline, scene design, drafting, continuation, revision, and consistency audit, with one primary mode per task.
- Scene cards and causality: each scene should change goal, information, relationship, resource, risk, or self-awareness, avoiding then chains.
- Pre-generation prediction: identify likely failure points, facts to protect, and the reader’s end-of-scene question.
- Post-generation backfill: update established facts, character knowledge, relationships, resources, suspense, and conflicts.
- Single-variable revision: diagnose causality, motivation, and information release before polishing sentences.

Boundaries

Use it for continuation, outlining, scene design, consistency checks, and prompt assembly; it does not replace writing itself. If no source text is available, it does not pretend to have read it and marks unknown chapters as unknown. Conflicting settings require evidence locations and repair options before choosing. Private drafts are used only for the current task and are not placed in public examples or logs.