长篇小说专业创作

SKILL内容创作
内容创作 未标注权限 许可证不明确
安装这个技能

Novel Writing 是一个以作者为最终决策者的长篇小说创作 skill:从创意、人物档案、故事提纲和章节卡,到分场景写作、对话与叙事修订、连续性审计和可视化管理;支持从指定网址或自有文本中抽象叙事与对话风格,生成语言风格模板,帮助优化小说文风。

一句话选型结论

按资料字段整理

长篇小说专业创作是内容创作类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类内容创作 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_8077ecd2/novel-writing-pro
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

内容创作
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_8077ecd2仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791728374
正文体量14.0 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
写文章与脚本

按主题产出初稿与分镜脚本。

改稿与润色

把草稿改顺、统一语气。

多平台文案改写

同一内容按平台改成长短不同的版本。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在内容创作场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
长篇小说专业创作当前 未标注权限 未标注 未标注 未标注 有
AI 视频生成 调用外部服务 GitHub 仓库 431.3 万 Codex、Claude、Cursor 有
视频帧 联网访问 ClawHub 市场 262.1 万 OpenClaw 有
Markdown 转换器 写入文件 ClawHub 市场 135.3 万 OpenClaw 有
哔哩哔哩全合一 接触敏感数据 ClawHub 市场 64.1 万 OpenClaw 有
Remotion video toolkit 命令行 浏览器控制 ClawHub 市场 63.9 万 OpenClaw 有

SKILL.md 全文

14.0 KB
模型实际读到的说明书原文,来自技能仓库

Version Update

This version adds permission-aware /novel:style-import for abstracting narrative and dialogue style from permitted reference text, layered character context for more focused drafting, and an optional eight-beat planning lens. Website source configurations are not bundled by default; Gutendex and Wikisource are not accessed unless a user supplies a custom configuration.

Novel Writing

For writers

This is not a one-click novel generator. It is an author-controlled writing partner that helps carry a story from a rough idea to a complete long-form manuscript through outlining, character development, chapter drafting, revision, continuity checks, and visual story tracking. The author keeps control of canon, voice, chapter boundaries, and every proposed change.

Core workflow

Execute these stages in order:

  1. Resolve the existing novel project or initialize one with scripts/init_project.py.
  2. Before /novel:chapter, run scripts/check_chapter_registry.py PROJECT_DIR. If it reports missing or stale metadata for existing manuscript files, stop drafting and offer scripts/scan_chapter_registry.py PROJECT_DIR --write; do not write a new chapter until the author accepts the scan or explicitly resolves the report. Then load the project-local working set from project-workflow.md. Read the previous chapter, relevant outline and character sections, voice guide, and continuity records; do not load the whole manuscript by default.
  3. Read all newly supplied fiction content before evaluating, questioning, outlining, drafting, or revising it.
  4. Apply questioning.md when development decisions remain. Ask 3–5 content-anchored questions unless the author skips questions or requests immediate output. Execute status, audit, review, and visualization requests directly.
  5. Apply research-workflow.md only when factual research can materially change plausibility or a story decision.
  6. Route the request to plotting.md, emotional-focus.md, character-arcs.md, character-context.md, drafting guidance below, dialogue-and-clarity.md, continuity.md, revision.md, or ai-writing-patterns.md. Apply the emotional spine, hierarchy, and pressure gates for every outline, draft, review, and revision.
  7. Preserve confirmed, inferred, and author-planned as separate certainty states. Merge no proposed state without explicit author confirmation.
  8. Draft or revise the author-selected chapter. Follow the author-maintained chapter plan; do not automatically partition the full novel into chapters or run an automatic multi-batch chapter pipeline.
  9. Save a completed chapter before review. Never silently overwrite a manuscript file; preserve the source unless the author requests an in-place edit.
  10. After craft revision, run the optional residual-pattern check in ai-writing-patterns.md before final continuity/state handling, then apply chapter-review.md and offer all three review modes. Resolve accepted revisions before updating structured state.
  11. Apply visualization-workflow.md. Extract candidate timeline, character, relationship, mystery, and clue changes; preview and merge them only through explicit author confirmation.
  12. Continue with the next author-selected chapter until the manuscript is complete, then perform a whole-project continuity and revision pass.

Operation routes

Operation Execute Load
/novel:new Initialize the project; establish premise, constraints, and initial story state. project-workflow.md, questioning.md, plotting.md, character-arcs.md
/novel:outline Develop the dramatic question, emotional spine, primary conflict, ranked threads, pressure ladder, causal structure, character change, author-chosen chapter plan, and chapter character card for the active cast. project-workflow.md, questioning.md, plotting.md, emotional-focus.md, character-arcs.md, character-context.md, continuity.md
/novel:chapter Run the chapter-registry preflight, then load the selected chapter card and exact active dossier sections, carry the scene emotional ledger through drafting, draft from confirmed hard constraints, save the selected chapter for review, and offer the fiction-safe residual-pattern loop. project-workflow.md, chapter-registry.md, emotional-focus.md, character-context.md, dialogue-and-clarity.md, continuity.md, ai-writing-patterns.md
/novel:review Run the selected manual, auto-review, or auto-review-and-revise mode, including the emotional spine, focus, pressure, and narrative voice/template-pattern dimensions. chapter-review.md, emotional-focus.md, dialogue-and-clarity.md, continuity.md, ai-writing-patterns.md
/novel:revise Preserve the source, run the emotional focus pass before ordered prose revisions, and propose one targeted second rewrite for accepted residual-pattern findings. questioning.md, emotional-focus.md, revision.md, dialogue-and-clarity.md, continuity.md, ai-writing-patterns.md
/novel:audit Run scripts/validate_story_data.py and scripts/audit_project.py; report registry, source-reference, and continuity contradictions before repairs. This remains continuity-first by default and does not run the AI-pattern review unless explicitly requested as a craft repair. chapter-registry.md, continuity.md, then revision.md only for approved repairs
/novel:visualize Preview candidate state changes, confirm a packet, and render selected views. visualization-schema.md, visualization-workflow.md
/novel:status Run scripts/project_status.py; summarize manuscript and continuity progress. project-workflow.md, continuity.md
/novel:style-import Search for and confirm a permitted reference work; accept website, author-owned file, public-domain text, or pasted text; cache local samples; run the style-import scripts; review qualitative abstractions; and optionally add the approved abstract template to the project voice guide. Refuse living-author imitation and protected-text copying. style-import.md, style-template-model.md, dialogue-and-clarity.md, character-context.md

CLI usage

  • Initialize: python3 scripts/init_project.py PROJECT_DIR --title TITLE [--language LANGUAGE].
  • Show status: python3 scripts/project_status.py PROJECT_DIR [--format text|json].
  • Validate structured state: python3 scripts/validate_story_data.py PROJECT_DIR.
  • Audit chapter metadata: python3 scripts/audit_project.py PROJECT_DIR. Run after ordinary schema validation and before repairing a chapter rewrite or split.
  • Scan legacy chapters: python3 scripts/scan_chapter_registry.py PROJECT_DIR [--write] [--status accepted|draft|review-pending|planned]. Preview first; --write creates continuity/chapters.json only when it has no populated registry and refuses to overwrite existing chapter records.
  • Check before a new chapter: python3 scripts/check_chapter_registry.py PROJECT_DIR [--json]. Treat a non-zero result as a drafting gate and offer the scan command before writing new prose.
  • Preview or merge an exact update packet: python3 scripts/merge_story_updates.py PROJECT_DIR UPDATE_FILE [--expected-sha256 HEX] [--dry-run].
  • Render an offline dashboard: python3 scripts/render_dashboard.py PROJECT_DIR [--output FILE] [--mode author|reader] [--types timeline,relationships,clues].

Use --expected-sha256 for every post-chapter merge. Present the exact digest with the complete candidate changes and merge only after the author confirms that digest.

Drafting guidance

  • Follow the confirmed chapter purpose, viewpoint, voice, chronology, knowledge boundaries, and style guide.
  • Give each scene a concrete goal, active resistance, meaningful turn, and consequential outcome.
  • Stage consequential conflict and relationship change as action and dialogue instead of summarizing the beat away.
  • Show emotion through bodily signal, behavior, perception, and selective interiority rather than relying on bare feeling labels.
  • Before drafting, read the selected chapter character card and exact active dossier sections according to character-context.md; never read every dossier or chapter card by default.
  • Before drafting a new chapter, require a passing chapter-registry preflight. If a legacy project has chapter files without records, pause and offer a preview plus explicit author-approved scan; after scanning, rerun the preflight and audit before proceeding.
  • Before drafting, load the chapter emotional question, ranked threads, cost of failure, and scene emotional ledger from emotional-focus.md. Every batch must carry entry pressure, immediate want, resistance, choice, emotional turn, and exit pressure.
  • Keep pre-draft continuity to hard constraints that can invalidate the chapter. Defer detail consistency such as fine speech habits, gestures, minor descriptions, background echoes, motifs, and callbacks to the first revision pass unless one is itself a hard constraint.
  • Derive ephemeral scene profiles for active speakers and maintain a knowledge ledger according to dialogue-and-clarity.md.
  • Bind every major clue or explanation to a present choice, cost, relationship, belief, or available option. If it changes none of these, defer, compress, or stage its human consequence before continuing.
  • If the project style/voice.md contains an approved style-import template, use it as the author's overall narrative and dialogue baseline. Treat quantitative signals as assistive checks only, never fixed sentence-length targets; preserve per-character voice, the name-blind test, and character-context rules, and follow the template's local copyright and permission boundaries.
  • Preserve deliberate ambiguity while keeping actor, action, speaker, time, place, and referent clear.
  • End where the chapter's change creates forward pressure; do not add a hook that contradicts its tone or resolution.

Fiction-safe AI-pattern residual loop

For /novel:chapter, /novel:review modes 2/3, and /novel:revise, preserve the first draft or source revision, read it at scene/paragraph scope, and apply the normal craft order before checking ai-writing-patterns.md. Record evidence, pattern ID, craft effect, possible intentional style, and uncertainty. Propose exactly one targeted second rewrite for accepted findings, keep the source separate, and then recheck canon, character/narrator voice, viewpoint access, and intentional irregularity. Findings and proposals remain non-final for human curation; never run a blanket humanization pass.

Precedence is character voice and scene pressure, then project/narrator voice and approved author samples, then the generic checklist. The review is not an AI detector, vocabulary blacklist, punctuation normalizer, detector-score gate, fixed-length/rhythm quota, or generic plainness mandate. /novel:status remains unchanged and /novel:audit remains continuity-first by default.

Author authority and safety

  • Never promote an inference, suggestion, or planned secret to confirmed canon without author confirmation.
  • Report contradictions before proposing repairs and preserve the author's chosen canon.
  • Treat brainstorming, summaries, review findings, and proposed prose as non-final until accepted.
  • Never silently overwrite manuscript files or structured story data.
  • Preserve the author's voice. Decline imitation of a living author's style and offer high-level craft traits instead.
  • Keep manuscript, research, outline, state, and visualization outputs inside the author's project.

常见问题

5 问
答案由本站依据条目字段整理
长篇小说专业创作怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_8077ecd2/novel-writing-pro。安装方式为未标注。装完刷新技能列表或重开客户端。

长篇小说专业创作支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

长篇小说专业创作权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

长篇小说专业创作收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

长篇小说专业创作用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。