Back to skill
Skillv1.0.0
ClawScan security
novel-generator 是一个中文爽文小说生成技能。用户只需提供一句话方向（如"写个都市重生爽文"），AI 代理即可自动完善提示词、规划大纲、逐章创作并输出为独立 Markdown 文件。 核心特性： 智能提示词生成：从一句话方向自动补全世界观、人设、冲突、爽点设计 分章节创作：每章 2000-3000 字，层层递进，章章有爽点 记忆系统：通过 .learnings/ 记录角色、地点、情节、世界观，确保故事前后一致 情节图解：关键战斗、人物关系、势力分布自动生成 Mermaid 图 失败记录：穿帮、矛盾、崩塌等问题自动记录，持续优化 多题材支持：都市、修仙、玄幻、重生、系统流、末世、科幻、游戏 兼容 Claude Code、Cursor、OpenAI Codex、GitHub Copilot 等所有支持 Agent Skills 的工具。 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 10, 2026, 7:22 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- This skill's files, instructions, and required permissions are consistent with its stated purpose (automated Chinese '爽文' novel generation) and do not request unrelated credentials or external installs.
- Guidance
- This skill appears coherent and limited to local novel-generation tasks, but before installing consider: (1) it needs file read/write permission in the skill folder — generated chapters and the .learnings/ memory will be written to disk; don't point it at a directory containing important unrelated files; (2) the init script supports --clean which will delete output/*.md and reset .learnings/ — back up any important outputs before running it; (3) the skill does not request network credentials or external endpoints, so it shouldn't exfiltrate secrets, but you should still review generated outputs for sensitive data you don't want persisted; (4) if you enable autonomous agent execution, remember the agent can create many files under output/ without further prompts — consider running interactively or in a sandboxed environment if you want tighter control.
Review Dimensions
- Purpose & Capability
- okName/description match what the package does: prompt expansion, outline planning, per-chapter generation, local memory (.learnings/) and Mermaid diagram output. The provided templates, examples, and init script are appropriate for a novel-generation skill.
- Instruction Scope
- okSKILL.md explicitly instructs the agent to read/write the .learnings/ files and write output/ markdown files; this is necessary for the stated continuity/remembering features. There are no instructions to access unrelated system paths, external endpoints, or to exfiltrate secrets.
- Install Mechanism
- okNo install spec or external downloads — instruction-only with a small helper shell script. No external URL downloads or package installs are present.
- Credentials
- okThe skill requires no environment variables, no credentials, and no config paths beyond its own directory (.learnings/ and output/). The requested filesystem access is proportional to maintaining story memory and outputs.
- Persistence & Privilege
- noteSkill will read and write files inside its skill directory (.learnings/ and output/). The included init script can delete output/*.md and resets .learnings/ when run with --clean (it uses rm -rf on output/*.md and overwrites learning files). This is expected for project initialization but users should be aware it will persist and can overwrite skill-local files.