zh-novel-writer - 批量生成引擎

批量生成网络小说章节。通过环境变读取 API keys 并调用外部 LLM API (ModelScope, Fyra, Ph8) 生成中文小说内容。 使用场景：用户给出大纲并要求"批量生成章节"、"写第X章到第Y章"、"后台静默写作"。 前置要求：需设置环境变量 NOVEL_MODELSCOPE_KEY 或 N...

What to consider before installing/using: - Privacy & billing: Using this skill will send your outline text, chapter excerpts, and prompt contents to whichever third-party LLM API you configure (ModelScope, Fyra, or Ph8). Do not include secrets or sensitive personal data in outlines or prompts. Be aware API usage may incur charges on the linked accounts. - API keys: The package metadata failed to list required environment variables, but the code requires at least one of NOVEL_MODELSCOPE_KEY, NOVEL_FYRA_KEY, or NOVEL_PH8_KEY. Only set keys for accounts you trust; prefer disposable or limited-billing keys if possible. - Inspect inputs/outputs: The script will read the outline file you point at and will write chapter files to the output directory. Ensure the outline path does not point to any sensitive files and run the script in a directory you control. - Run safely: If you want to be cautious, run the script in an isolated environment (container or VM) and monitor network calls. You can also review and/or modify the CONFIG dictionary in the script to verify endpoints and headers before using. - Packaging correctness: The mismatch between registry metadata and the SKILL.md (env vars not declared in metadata) is likely a packaging oversight — consider asking the publisher for clarification or updating metadata before broad deployment. If you need higher assurance, request the publisher to declare required env vars in registry metadata, provide a signed release, or supply usage examples showing how keys are scoped and billed.

Type: OpenClaw Skill Name: zh-novel-writer Version: 1.0.3 The skill bundle is a legitimate tool designed to batch-generate novel chapters using external LLM APIs (ModelScope, Fyra, and Ph8). The Python script `scripts/batch_generate.py` correctly implements its stated logic, including environment variable handling for API keys, retry mechanisms for rate limits, and local file storage for generated content. No evidence of data exfiltration, unauthorized network calls, or malicious prompt injection was found; the tool's behavior is entirely consistent with its documentation.