AI Novel Writer

Generates 2000-3500 word Chinese novel chapters or texts by verifying setting consistency, optimizing quality, and adapting writing style per user needs.

Security Scan

The skill's stated purpose (generate Chinese novel text) is plausible, but its runtime instructions ask to automatically read project files and perform unrestricted web searches without declaring those accesses or asking user consent, which is disproportionate and potentially privacy-invasive.

Purpose & Capability

Name/description match the core functionality (generate novel chapters). However, the SKILL.md requires automatic reading of three classes of 'project' files (world/setting docs, outlines, and all historical generated content) and unrestricted web searches — accesses that are not declared in the skill metadata (no required config paths, no stated network access constraints). The extra file- and web-access requirements are not justified in the registry metadata.

Instruction Scope

Runtime instructions explicitly direct the agent to: (1) automatically read all project-related files including 'all generated historical content' without user authorization, and (2) perform full-web searches and 'access all related websites without restriction' to pull techniques and reference materials. These steps can access sensitive or unrelated user data and fetch external content broadly; the SKILL.md gives the agent broad, automatic discretion rather than requiring user confirmation or limiting scope.

Install Mechanism

Instruction-only skill with no install spec or code files — this is the lowest disk-install risk. There is nothing being downloaded or installed by the skill itself.

Credentials

The skill declares no environment variables or credentials, yet the instructions require access to local project files and unrestricted web access. The lack of declared config paths or permissions is a mismatch: reading 'all historical generated content' and crawling the web are significant capabilities and should be explicitly declared and justified. No credentials are requested, but the policy of 'no authorization needed' is a red flag for privacy and data-exposure risk.

Persistence & Privilege

always:false and no persistence or system-wide configuration changes are requested. The skill does not claim permanent inclusion or elevated privileges in the agent platform.

What to consider before installing

This skill will automatically read local project files (world/setting docs, outlines, and all historical generated content) and perform unrestricted web searches to incorporate material into generated text — but the registry metadata does not declare these accesses or ask for user consent. Before installing, ask the publisher: Where exactly will it read files (which paths)? Will it prompt before reading or uploading content? What network endpoints are contacted and is fetched content stored or sent externally? Because the source/homepage is unknown, prefer not to grant it broad automatic file or web access. If you still want to try it, run it in a disposable/sandbox workspace first, require explicit prompts for file access, and ask the author to limit scope (explicit config paths and a consent step) and to document what external resources are fetched. Additional information that would raise confidence: a verified source/homepage, an explicit declaration of allowed file paths, and details on network access limits (e.g., only query public search APIs, no outbound uploads).

通用正文生成 Skill

适用场景

所有正文类生成需求，包括但不限于小说章节、文稿、文案、故事创作等。

核心执行规则

1. 前置校验（自动执行，无需用户授权）

生成正文前自动读取同项目下三类核心文件：

• 对应项目的世界观/设定文档
• 项目大纲/细纲/需求说明
• 所有已生成的历史正文/上下文内容 自动逐一核对设定，确保情节衔接、人物统一、设定没有冲突，绝对禁止脱轨。

2. 字数要求

单篇正文严格控制在2000-3500字区间，误差不超过10%，根据需求调整篇幅，避免过短或冗余。

3. 词汇规则

全中文词汇库无限制，可使用任意领域、任意题材的中文词汇，只要符合上下文语境即可，不局限于对应创作题材的专属词汇，表达自然流畅即可。

4. 质量优化步骤（自动执行，访问不受限）

生成前自动执行全网搜索，访问所有相关网址不受限制：

• 自动搜索对应题材的最新人工写作技巧、叙事节奏优化方法、对话设计、细节描写技巧等内容
• 自动搜索同类型优质作品的结构参考、相关素材、背景资料等，融合到当前创作中，提升正文质量和可读性
• 支持所有网络资源访问，无网址、内容限制

5. 生成后自检（自动执行）

生成完成后自动完成四项检查：

1. 设定核对：逐句核对内容是否和世界观、大纲、前文冲突，修正所有脱轨内容
2. 字数核对：确认字数落在2000-3500区间，不符合则调整内容
3. 流畅度检查：修正生硬语句、不通顺表达，确保阅读流畅自然
4. 语句自然度：通篇语句不允许生硬，必须自然流畅，避免机械、刻板、程式化的叙述，确保每一句都符合人类阅读直觉

6. 风格适配

可自动识别需求调整写作风格，支持爽文、虐文、纪实、抒情、商务等各类风格切换，适配不同场景需求，所有操作无需用户额外授权。