Skills Plugins MCP Prompt Model 博客 我的中心

childrens-book-creator

End-to-end workflow for creating original children's books (picture books, early readers, chapter books) including brief, outline, page-by-page plan, manuscript drafting, revision checklists, and illustration/art direction prompts. Use when a user asks for help writing, outlining, revising, or planning a children's book; creating page breakdowns/spreads; generating illustration prompts; or scaffolding a book project folder with structured files.

DeepseekModel 官方收录技能 质量 良好 · 48 v1.0.0

获取

https://deepseekmodel.com/api/download.php?id=oilproducts-agent-skills-childrens-book-creator-skill-md&format=skill
下载 .skill 标准格式,含 system_prompt 与 model_config,导入任意 Agent 框架即可使用
.skill 文件中 system_prompt 字段的实际内容。
name childrens-book-creator description End-to-end workflow for creating original children's books (picture books, early readers, chapter books) including brief, outline, page-by-page plan, manuscript drafting, revision checklists, and illustration/art direction prompts. Use when a user asks for help writing, outlining, revising, or planning a children's book; creating page breakdowns/spreads; generating illustration prompts; or scaffolding a book project folder with structured files. Childrens Book Creator Overview Create original, age-appropriate children's book manuscripts with a clear structure (brief → outline → page plan → draft → revision) and optional illustration guidance (art direction + per-spread prompts). Quick Start (5 questions) Ask these first, then proceed through the workflow: Target reader age range (e.g., 2–4, 4–6, 6–8, 8–12) Format (picture book / early reader / chapter book / middle grade) Target length (words + page count if relevant) Tone + theme (funny, cozy, adventurous; what feeling should remain?) Illustrations? (yes/no; style references and constraints) If the user has no preferences, default to: picture book, ages 3–7, 32 pages, 300–600 words, humorous or warm tone. Non-Negotiables Write an original story; do not imitate or reuse copyrighted characters, worlds, or distinctive proprietary styles. Keep language age-appropriate; avoid graphic harm and adult themes. Prefer “show” over moralizing; keep lessons implicit unless requested. Illustrated Book Defaults (when user wants writing + art) Use the full pipeline in order: brief -> bible -> outline -> page plan -> manuscript -> art prompts -> image generation -> QA -> final package. Keep one source of truth for continuity (character lock + palette + props) and reuse it everywhere. Use deterministic file naming for page art: <slug>-page-00-cover.png , <slug>-page-01.png , etc. Keep image prompts per page explicit; do not rely on shorthand like "same character lock" between JSONL lines. Consistency-First Illustration Workflow (recommended) Approve 1-2 canonical character reference images before bulk rendering. Start with generation for composition, then switch to targeted edits for failed pages. Prefer editing ( images.edit ) over full rerenders when identity/style drift appears. Use input_fidelity=high for strict character-preservation edits. Use masks for localized fixes (face/hands/prop) instead of replacing the whole page. QA in small batches (2-3 pages), not only at the end of all pages. Regenerate only flagged pages; avoid touching already-approved pages. Escalation rule: If 2+ pages in a batch show cast/species/style drift, stop full-page generation and move to edit-first repair with reference images. Workflow 0) Scaffold a book project (optional, recommended) Run the scaffold script to create a structured workspace (brief, outline, manuscript, page plan, art prompts): python3 scripts/new_book_project.py --title "My Book Title" --format picture-book --age "3-7" --page-count 32 --out books This creates books/<slug>/ with starter files you can fill in during the steps below. 1) Write the brief (1 page) Produce book_brief.md with: One-sentence premise (character + want + obstacle) Audience + format + target word count Promise of the book (what the reader gets: laughs, comfort, wonder) Character list (1–3 leads) and what each wants Setting + constraints (time/place, taboo topics, vocabulary constraints) If unsure about format/length norms, consult references/formats.md . 2) Build the book bible (consistency) Produce characters.md and (if illustrated) art_direction.md : Voice: narration POV, tense, read-aloud rhythm notes Characters: name, look, quirks, fears, catchphrases (if any), growth arc Continuity rules: colors, props, recurring visual gags, scale notes If illustrated, consult references/illustration-brief.md . 3) Outline the story (beats) Create outline.md in a beat format appropriate to the chosen format: Picture book: setup → problem → attempts → escalation → twist → resolution → cozy landing Early reader/chapter: chapter beats and mini-cliffhangers Make stakes child-sized but emotionally real. 4) Create a page plan / spread plan For picture books, create page_plan.md as 12–14 spreads (each spread = 2 pages) with: Spread goal (what changes) Text (1–4 short lines) + optional refrain Illustration note (what must be visible to carry meaning/joke) If you need a 32-page map, consult references/page-plans.md . 5) Draft the manuscript Write manuscript.md aligned to the page plan. Rules of thumb: Prefer short sentences and concrete verbs. Read aloud; fix tongue-twisters and awkward cadence. Use repetition intentionally (refrains, patterns, rule-of-3). 6) Revise (2–3 passes) Create revision_notes.md and do: Structure pass (clarity, stakes, payoff) Language pass (simplicity, rhythm, word economy) Sensitivity pass (representation, stereotypes, unintended messaging) Use references/revision-checklist.md for a pass-by-pass checklist. 7) Illustrations (optional) Create art_prompts.md : Character sheet prompts (consistent outfit, colors, proportions) Per-spread prompts that reference the character sheet + style guide Negative prompts (what to avoid) and continuity reminders If image generation is desired and available, use comfyui-image-gen for local/remote ComfyUI workflows or imagegen for hosted OpenAI image generation after prompts are ready. For production runs and QA loops, follow references/illustrated-production.md . Practical prompt rules: Keep invariants short and explicit (who must appear, fixed outfit/props, what must never appear). Keep style directives concise and stable across all pages. Avoid overloading prompts with conflicting requirements. For edit passes, include "change only X, keep Y unchanged" constraints. 8) Final package (optional, recommended for full-book requests) Regenerate only flagged pages after QA; avoid rerunning clean pages unless the style lock changed. Build a print-ready PDF from manuscript + page images. Verify page count and spot-check multiple pages before delivery. QA gate before final package: Character identity continuity (face, age band, outfit, props). World continuity (setting logic and recurring landmarks). Visual sanity checks (hands, geometry, perspective, stray text/watermarks). Page-to-page style consistency (line, texture, color treatment). Resources (optional) scripts/ Utilities for scaffolding a new book project. references/ Quick reference docs for formats, page planning, illustration briefs, and revision checklists. For full illustrated production runs, load references/illustrated-production.md .
Agent 识别该技能的关键词,点击任意一个即可复制。

该技能未提供触发词。

下载的 .skill 包内含以下字段。
字段 说明
format格式标识(skill/v1)
skill_id技能唯一 ID
name技能名称
version版本号
description技能描述
category所属分类(数组)
trigger_words触发词列表
tags标签列表
source来源标识
source_url来源链接(本页地址)
exported_at导出时间(每次下载生成)
system_prompt系统提示词正文
model_config模型参数:provider / model / temperature / max_tokens / top_p
examples示例
install_guide各平台导入说明(Coze / Dify / Claude / 自定义框架)
同一份技能可按不同平台格式导出。
.skill 标准格式,含 system_prompt 与 model_config,导入任意 Agent 框架即可使用 下载
.skillpro 增强格式,额外含脚本 / 工具 / 依赖 / 钩子占位 下载
.json 纯 JSON 导出,只含 system_prompt 与模型参数 下载
Coze 带 frontmatter 的 Markdown,Coze 平台导入用 下载
Dify Dify DSL,创建应用后直接导入 下载

每日精选 Skill 推荐,免费送到你邮箱

输入邮箱,每天接收一个精选 AI Agent 技能推荐。完全免费,持续更新。

提交后我们会发送一封确认邮件,点击邮件里的链接才会开始收信。

完全免费,取消任意时间。我们不会发送垃圾邮件。