Overview

A CI/CD-style chapter pipeline for serialized web novels, running inside your AI agent. NovelOps is an Agent Skill: a SKILL.md scheduling protocol, 10 role prompts, 15 dynamic track templates, and 3 Python tool scripts. After installing any AI Agent that supports Skill (TRAE, Claude Code, etc.), say "Write a chapter," and the Agent will autonomously complete the entire production cycle — writing, quality check, rewriting callback, status placement — stopping only at 4 key decision points for your confirmation. It brings the CI/CD discipline of software engineering into the serialized novel: access control passes or you don't proceed to the next chapter, all status is recorded as plain text, and if a conversation breaks, it resumes from the breakpoint. One click generates 100,000 words, and the ending inevitably involves setting drift, forgotten foreshadowing, missing subplots, and repeated tropes. NovelOps tracks these issues chapter by chapter with 15 dynamic tracks: foreshadowing over 20 chapters is not recovered and marked in yellow; over 40 chapters are marked in red; formulas have cooldown periods; if not expired, they must be replaced with significant variants; emotions are forcibly adjusted in the same direction for 3 consecutive chapters; each subplot must appear every 8 chapters. Sentence length is uniform, vocabulary is sparse, and writing style tends to be similar. Before writing, NovelOps takes target feature cards from the sample library and aligns them. During quality checks, metrics.py is used to measure σ² (sentence length variance), TTR (vocabulary richness), EGR, Pσ², and character language fingerprints; Every 10 chapters, homogeneity checks and style stability checks are conducted (when SC < 0.6, the stable chapters are used as the benchmark to rewrite drifting chapters). Chinese online literature platforms have clear rules for AI content. The quality inspection gate has built-in AI ≤ 30%, originality ≥ 70%, plagiarism check ≤ 2 places, redline word scanning, and after final review, a level 3 sensitive word check before exporting. 10 Role Agents in Sequential Relay: Topic Market → Topic Selection → Core Idea Card → Material Preparation → Outline → Characters → Rhythm → Operations Optimization → Creative Disruption. Among these, the Core Idea Card is the most critical step. Five settings must be finalized in collaboration with the author: City (irreplaceability of the core contradiction), Industry, Core Capability (including costs and limitations), Core Contradiction, Creative Differentiation (comparison with the specific blockbuster). Any ambiguity is rejected; after all pass, the index is written into the entire process—AI has no right to modify it. There is an upgrade mechanism for rejection: the first partial revision, the second full chapter rewrite, the third trigger the creative disturbance agent to inject variables, the fourth upgrade with manual review.

Overview (machine translated)

This page was translated automatically. Technical details may differ from the original.

Recommended Tools

Try a different keyword or remove a filter.

Install