novel creator
v3.0.1为中文长篇小说、连载网文、章节样稿、已有草稿续写提供一套完整且内聚的创作流程：分流新建/续写/救稿任务，维护 plan/ 与 memory/ 记忆系统，在写正文前先学习指定题材的文风笔触，再按节奏持续产出章节，并在同一技能内完成文风校准、题材适配、连载维护与必要的创作研究。适用于用户要求写小说、开书、搭大纲、出样...
⭐ 0· 50·0 current·0 all-time
by@km1001
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (novel creation, planning, style preheat, sustaining a serial) match the included assets, references, and scripts. The two scripts (init and wordcount) and many reference/example texts are appropriate for the stated purpose. Minor inconsistency: SKILL.md and README mention a PowerShell init script (scripts/init-novel.ps1) but the manifest only contains scripts/init-novel.sh; this is likely an omission rather than malicious, but worth noting.
Instruction Scope
SKILL.md explicitly instructs reading local examples/references and creating/updating plan/, memory/, output/, manifest.json, and running the init/wordcount scripts. All file reads/writes stay within the workspace and are consistent with the described functionality. There are no instructions to read unrelated system configuration, shell history, or to send data externally. The workflow does include an option to --clean (which will remove plan/memory/output/manifest after backing them up) — users should confirm before invoking clean.
Install Mechanism
No install spec and no external downloads — this is an instruction-first skill with bundled scripts and templates. That is low risk. The included Bash init script and Python wordcount script operate on the local filesystem; there are no remote URLs, extract steps, or package installs.
Credentials
The skill declares no required environment variables, no credentials, and no config paths. The SKILL.md and code read/write only local files under the working directory (plan/, memory/, output/, backup/). There are no hidden environment accesses or secret usages.
Persistence & Privilege
always is false. The skill will create and update files under the current working directory as part of its normal operation (expected for a project/workspace tool). Autonomous invocation (disable-model-invocation: false) is the platform default; combined with the skill's local-file behavior this is reasonable. The only elevated filesystem action to note is the optional --clean flow which deletes workspace directories (but the script performs a backup step first).
Assessment
This skill appears to do exactly what it says: it reads local examples and reference texts, creates/upates plan/, memory/, output/, and manifest.json, and includes two simple scripts (init and wordcount). Before running: 1) Be aware the init script will create files in the current directory and, when used with --clean, will back up then delete old plan/memory/output/manifest — so run it in the intended folder and inspect the script first. 2) The skill does not request network access or credentials, but it will read any files placed under the workspace (so avoid putting secrets in that folder). 3) There is a small manifest inconsistency: README/SKILL.md mention a PowerShell init script that isn’t present; you may want to inspect the shipped Bash script (scripts/init-novel.sh) to ensure it matches your expectations. If you want extra assurance, open the two scripts (init-novel.sh and check_chapter_wordcount.py) and the templates in assets/init before invoking the skill.Like a lobster shell, security has layers — review code before you run it.
latestvk9731737zjxb2k891vhw510m4x84ssng
License
MIT-0
Free to use, modify, and redistribute. No attribution required.