What problem does it solve?
This Skill removes common quality failures in long-form AI web novels—rule violations, inconsistent character voice/honorifics, timeline/number mismatches, weak hooks/openings, and low mobile readability—while steadily improving each episode in sequence.
Core Features & Use Cases
- Parallel 4-agent diagnosis, then staged correction: runs rule-checking, story/timeline/number analysis, platform optimization, and character “alive” enhancement before applying revisions.
- Self-loop, episode-by-episode execution: automatically advances through the next unfinished episode without waiting for user input, updating progress and learning logs.
- Project-driven guardrails and custom axes: loads project-specific settings from novel-config.md (required) to apply platform rules, banned patterns, voice/title rules, numeric/time verifications, and optional custom diagnostic axes.
- Strict validation and retry: verifies the corrected episode (including final timeline/number checks) and only revises when critical issues remain; targets continuity with the previous two episodes.
Quick Start
Run the polish workflow from the next unfinished episode for your current project by typing: /polish.