novel-writing

AI长篇网文创作技能包。用于解决长篇网络小说创作中的核心痛点：上下文丢失、文风不一致、设定冲突、节奏失控、多线混乱、质量不稳、读者反馈无法内化。触发场景包括：开始新书、规划大纲、撰写章节、管理伏笔、检测冲突、读者反馈分析、批量创作质量控制。

Security Scan

This is a locally focused novel-authoring skill that manages project files, tracks state, and performs simple checks; it does not request credentials or perform network access and its requirements match its stated purpose.

Purpose & Capability

Name/description (long-form novel authoring: outline, chapters, foreshadowing, conflict checking, feedback tracking) match the included content and CLI script: the code creates/reads/writes files under novel/<project>, keeps state.json, and provides conflict/foreshadowing/feedback utilities. No unrelated credentials or binaries are requested.

Instruction Scope

SKILL.md describes AI behaviors like '自动召回相关设定+前5章摘要, 按文风指南输出' which relies on the agent reading the local project files; the provided script implements local file I/O, outline/conflict/feedback tracking, and basic checks. There is no network I/O or access to system paths outside the novel/ workspace. Note: some SKILL.md wording implies automated stylistic transformation by an LLM — the script itself only manages files and checks.

Install Mechanism

No install spec (instruction-only skill) and no external downloads; a single benign Python CLI file is included. Nothing is fetched from the network or extracted during install.

Credentials

The skill does not request environment variables, credentials, or config paths. The code does not read sensitive environment variables or external auth tokens.

Persistence & Privilege

always:false and normal autonomous invocation settings. The skill only reads/writes files inside its own novel/<name> workspace and does not modify global agent configuration or other skills.

Assessment

This skill appears internally consistent and operates locally: it creates and modifies files under a novel/<project> directory and does simple conflict/foreshadowing checks. Before installing/running: (1) review the included scripts/novel.py to confirm you are comfortable executing a Python CLI from an unknown source; (2) run it from a directory where writing novel/<name>/ is safe (it will create and update files and state.json); (3) note there are no network calls or secret requests, so credential exposure risk is low; (4) understand that 'automatic recall' and style enforcement are responsibilities of the agent/LLM using these files — the script itself only manages and checks files; (5) consider keeping backups of important work and inspect/validate the code if you need stronger guarantees about behavior or want to remove/modify any automatic file writes.

AI长篇网文创作技能包，解决长篇连载中的7大痛点。

核心架构

novel/
├── settings/          # 世界观设定档案
│   ├── world.md       # 世界观总览
│   ├── characters/    # 角色声线卡
│   ├── geography/     # 地理/势力
│   └── rules.md       # 世界规则
├── chapters/          # 章节档案
│   ├── outline.md     # 大纲
│   └── arc-N/         # 分卷
├── tracker/           # 追踪器
│   ├── foreshadowing.md  # 伏笔追踪
│   ├── conflicts.md    # 冲突记录
│   └── feedback.md    # 读者反馈
└── state.json          # 当前写作状态

工作流程

1. 开书初始化 novel init <书名>

创建项目结构，加载默认模板。

2. 设定录入 novel setting <类型> <内容>

录入角色/地理/规则设定，自动检测与现有设定的冲突。

3. 大纲规划 novel outline <章节范围> <一句话概要>

生成章节大纲，标注伏笔埋入点。

4. 撰写章节 novel write <章号> <正文>

自动召回相关设定+前5章摘要，按文风指南输出。

5. 伏笔检测 novel check <章号>

检查本章埋入的伏笔是否已在后续章节收回。

6. 冲突检测 novel conflict <新内容>

检测新内容是否与已有设定冲突。

7. 反馈入库 novel feedback <章节> <反馈内容>

将读者反馈存入追踪器，写相关章节时自动加载。

伏笔追踪规则

每埋一个伏笔必须登记：

- [章X] 事件A → 预期回收章Y

连续3章未回收则预警。

文风指南

在 references/style-guide.md 定义：

• 叙事语气（简洁/华丽/悬疑/热血）
• 句式偏好（长句/短句比例）
• 禁忌词表

状态文件

state.json 记录：

• 当前卷/章
• 已埋未收伏笔数
• 本日产出字数
• 最后修改时间

每次 novel write 后自动更新。