← 返回 Skills 市场
philipstark

Bestseller Writer

作者 PhilipStark · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
325
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install bestseller-writer
功能描述
Turn a shower idea into a full best-seller manuscript in one command. Multi-agent pipeline — Planner → Character Designer → Chapter Writers (parallel) → Edit...
使用说明 (SKILL.md)

Bestseller Writer 📚

One idea → Full manuscript → Amazon KDP listing. Fully autonomous.

Turn a shower thought into a publishable book. 5-stage multi-agent pipeline handles everything: story structure, characters, all chapters (written in parallel), editorial pass, and a complete Amazon KDP publishing package.

No writing experience needed. No blank page. Just your idea.


What You Get

Output Description
MANUSCRIPT.md Complete book (50-80k words, 25-35 chapters)
kdp_package.md Title options, Amazon description, 7 keywords, BISAC categories, pricing strategy, cover brief + Midjourney prompt
plan.md Full story/argument structure with chapter-by-chapter outline
characters.md Deep character profiles with arcs (fiction) or authority framework (non-fiction)
editorial_memo.md Professional developmental edit notes

Quick Start (CLI)

Install dependency

cd skills/bestseller-writer/scripts
npm install

Generate a book

# Thriller
node generate.js --idea "A detective discovers evidence of her own murder" --genre thriller

# Self-help  
node generate.js --idea "How to build a $10k/month business in 90 days using AI" --genre self-help

# Romance
node generate.js --idea "Two rival food truck owners forced to share a festival spot" --genre romance

# Business/non-fiction
node generate.js --idea "Why most startups fail at hiring and how to fix it" --genre business

Options

--idea, -i     Your book concept (required)
--genre, -g    Genre: thriller, romance, self-help, business, memoir, fantasy, nonfiction
--output, -o   Output directory (default: ./book-output/[slug])
--chapters     Number of chapters (default: 25)
--batch        Parallel chapter batch size (default: 4)
--planner      Model for Planner + Editor (default: claude-opus-4-5)
--writer       Model for writing agents (default: claude-sonnet-4-5)

Environment

export ANTHROPIC_API_KEY=sk-ant-...

Agent-Orchestrated Mode

If you're running inside an OpenClaw agent, the agent follows this pipeline directly using sessions_spawn. Trigger with:

"Write me a book about [idea]" or "Generate a bestseller about [idea]"

The agent will:

  1. Create output directory at ~/Desktop/books/[slug]/
  2. Run all 5 stages sequentially (chapters in parallel batches)
  3. Deliver kdp_package.md as the final deliverable
  4. Report word count and publish checklist

Pipeline Architecture

💡 Your Idea
     │
     ▼
┌─────────────────────────────────────┐
│  Stage 1: PLANNER (Opus)            │
│  Market positioning, title options, │
│  full chapter-by-chapter outline,   │
│  comparable titles, pitch           │
└──────────────────┬──────────────────┘
                   │
                   ▼
┌─────────────────────────────────────┐
│  Stage 2: CHARACTER DESIGNER        │
│  (Sonnet)                           │
│  Deep profiles, arcs, voice guide   │
│  (fiction) or authority/reader      │
│  avatar framework (non-fiction)     │
└──────────────────┬──────────────────┘
                   │
         ┌─────────┴─────────┐
         ▼                   ▼
  ┌─────────────┐     ┌─────────────┐
  │ Chapters    │ ... │ Chapters    │  ← Parallel batches
  │ 1-4 (Sonnet)│     │ 21-25       │    (4 at a time)
  └──────┬──────┘     └──────┬──────┘
         └─────────┬─────────┘
                   ▼
┌─────────────────────────────────────┐
│  Stage 4: EDITOR (Opus)             │
│  Consistency, pacing, voice,        │
│  opening/ending assessment,         │
│  marketability score                │
└──────────────────┬──────────────────┘
                   │
                   ▼
┌─────────────────────────────────────┐
│  Stage 5: KDP PACKAGER (Sonnet)     │
│  Title, description, keywords,      │
│  categories, pricing, cover brief,  │
│  Midjourney prompt, launch checklist│
└─────────────────────────────────────┘
                   │
                   ▼
        📦 MANUSCRIPT.md + kdp_package.md

Genre Support

Genre Framework Avg Length
Thriller/Mystery Save the Cat beats + 3-act 70-80k words
Romance Meet cute → Black moment → HEA 60-70k words
Fantasy/Sci-Fi World + Hero's Journey 80-100k words
Self-Help Problem → Framework → Transformation 40-50k words
Business Insight → Evidence → Application 45-55k words
Memoir Chronological + thematic arc 55-70k words

Revenue Model

Amazon KDP passive income per book:

  • Average: $50-500/month
  • Stack 10 books: $500-5,000/month
  • Stack 30 books: $1,500-15,000/month

Launch strategy (built into kdp_package.md):

  1. Publish at $0.99 → enroll KDP Select
  2. Day 1-4: Paid launch, collect reviews
  3. Day 5-9: Free promo (spikes ranking)
  4. Day 10+: Raise to $2.99-$4.99 (70% royalty)
  5. Repeat with next book

Niche targeting (highest ROI):

  • Cozy mysteries with recurring characters (series = repeat buyers)
  • Self-help for specific professions (accountants, nurses, teachers)
  • Business books for specific industries
  • Local history/interest books (low competition)

Cost Estimate

Stage Model API Cost (approx)
Planner Opus ~$0.15
Characters Sonnet ~$0.05
25 chapters Sonnet × 25 ~$0.75
Editor Opus ~$0.20
KDP Package Sonnet ~$0.05
Total ~$1.20 per book

One book costs ~$1.20 to generate and can earn $50-500/month. ROI is infinite.


Agent Instructions (for OpenClaw)

When the user asks to write a book or generate a manuscript:

Step 1 — Collect Info

Ask for (or infer from context):

  • The idea (required)
  • Genre (infer if obvious, otherwise ask)
  • Fiction or non-fiction

Step 2 — Run Pipeline

Create output dir: ~/Desktop/books/[slug]/

Spawn agents in sequence using sessions_spawn with runtime="subagent":

Agent 1 — Planner: Task: Full planning prompt (see scripts/generate.js buildPlannerPrompt) Save output to: plan.md

Agent 2 — Character Designer: Task: Character/authority prompt with plan as context Save output to: characters.md

Agents 3-N — Chapter Writers (parallel, max 4 at once): Task: Chapter writing prompt with plan + characters + prev chapter ending Save output to: chapter_NN.md

Agent N+1 — Editor: Task: Editorial pass prompt with plan + sample chapters Save output to: editorial_memo.md

Agent N+2 — KDP Packager: Task: KDP package prompt with plan + editorial Save output to: kdp_package.md

Step 3 — Assemble

Concatenate all chapter_NN.md files into MANUSCRIPT.md

Step 4 — Deliver

Send user:

  • Word count
  • Path to MANUSCRIPT.md
  • Key items from kdp_package.md (chosen title, pricing, first keyword string)
  • Next steps for publishing

Troubleshooting

Chapters are too short? The writer agents are prompted for 2,000-2,500 words. If output is shorter, re-run that specific chapter with: "This chapter is too short. Expand to at least 2,000 words, adding more scene depth, dialogue, and sensory detail."

Voice is inconsistent? The editor stage catches this. After the editorial memo, re-run any flagged chapters with the specific feedback.

KDP keywords not relevant? Edit kdp_package.md keywords manually using Google Keyword Planner or Publisher Rocket to verify search volume before uploading.

Want a series? After Book 1 is done, pass plan.md + characters.md into a new run with --idea "Book 2: [continuation]". Characters and world are already built.

安全使用建议
Key things to check before installing or running: 1) The code and docs require ANTHROPIC_API_KEY, but the registry metadata incorrectly lists no required env — don't trust the registry listing; provide only a dedicated Anthropic key with limited billing controls. 2) The package uses @anthropic-ai/sdk from npm — review package.json and package-lock (if present) before npm install and consider installing in an isolated environment. 3) Inspect scripts/generate.js for any uses of child_process.execSync or other shell operations (generate.js imports execSync); confirm what commands (if any) are executed so you aren't surprised by shell calls. 4) Understand where files will be written (script default is ./book-output/[slug], but SKILL.md references ~/Desktop/books/[slug]); run it in a sandbox or set --output explicitly to a safe directory. 5) If you plan to run agent-orchestrated mode, know that the agent will spawn multiple model sessions (normal for a pipeline) and will send your prompts/data to Anthropic — don't include any secrets or private data in the idea prompt. 6) Ask the publisher to correct the registry metadata to declare ANTHROPIC_API_KEY as a required credential and to clarify the exact filesystem paths and any shell commands the script may run.
功能分析
Type: OpenClaw Skill Name: bestseller-writer Version: 1.0.0 The 'bestseller-writer' skill is a multi-agent pipeline designed to generate full-length book manuscripts and Amazon KDP publishing packages using the Anthropic API. The provided Node.js script (generate.js) and the agent instructions (SKILL.md) align perfectly with the stated purpose of orchestrating LLM sub-agents for planning, character design, writing, and editing. There is no evidence of data exfiltration, malicious execution, or obfuscation; the script handles the user's API key locally and writes output to the filesystem as expected for a content generation tool.
能力评估
Purpose & Capability
The skill name/description (autonomous book generator + KDP package) matches the code and SKILL.md: generate.js orchestrates multi-stage prompts and produces manuscript and KDP files. However, registry metadata lists no required environment variables or credentials while both SKILL.md and scripts/generate.js explicitly require ANTHROPIC_API_KEY. That omission in the declared requirements is an inconsistency.
Instruction Scope
SKILL.md instructs the agent and user to provide an Anthropic API key, to run npm install, and (when run by an agent) to create output at ~/Desktop/books/[slug]/ and run the 5-stage pipeline. The runtime instructions are scoped to writing generated content and orchestrating model calls; they do not instruct reading arbitrary system files or exfiltrating data. The document does advise use of sessions_spawn when run inside an agent (expected for multi-agent orchestration).
Install Mechanism
This is instruction‑plus-code (no registry install spec). The included scripts/package.json depends on @anthropic-ai/sdk from npm — a standard public registry dependency. The install flow is 'npm install' in the scripts folder (no arbitrary downloads or remote extract). This is moderate, expected risk for a script that uses a cloud SDK.
Credentials
Functionally the skill needs one cloud credential (ANTHROPIC_API_KEY) to call Anthropic models — that is proportionate to the stated purpose. But the registry metadata claims 'Required env vars: none' and 'Primary credential: none', which is inaccurate and misleading. Any missing or incorrectly declared credential requirements are a red flag because users may not realize they must provide a secret key. Also the code reads process.env for the Anthropic key; no other secrets are requested — so the scope of secrets is limited but the metadata mismatch must be addressed.
Persistence & Privilege
The skill does not request elevated platform privileges and 'always' is false. Its normal behavior is to write output files to a specified output directory (script default is ./book-output/[slug]; SKILL.md suggests an agent path at ~/Desktop/books/[slug]/). It does not claim to modify other skills or system-wide settings. That is expected for a content-generation script.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install bestseller-writer
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /bestseller-writer 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release: full autonomous book pipeline, KDP packager, CLI script
元数据
Slug bestseller-writer
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Bestseller Writer 是什么?

Turn a shower idea into a full best-seller manuscript in one command. Multi-agent pipeline — Planner → Character Designer → Chapter Writers (parallel) → Edit... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 325 次。

如何安装 Bestseller Writer?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install bestseller-writer」即可一键安装,无需额外配置。

Bestseller Writer 是免费的吗?

是的,Bestseller Writer 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Bestseller Writer 支持哪些平台?

Bestseller Writer 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Bestseller Writer?

由 PhilipStark(@philipstark)开发并维护,当前版本 v1.0.0。

💬 留言讨论