Token导航 LogoToken导航TokenDH.com
开发权限需确认github未标认证来源可访问许可证需确认审计通过

prose-writing散文写作

Agent Skill

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

总安装

2,230

周安装

92

GitHub Stars

147

下载量

729
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:prose-writing(散文写作)
来源仓库:https://github.com/haowjy/creative-writing-skills
仓库路径:skills/prose-writing
安装命令:
npx skills add https://github.com/haowjy/creative-writing-skills --skill prose-writing
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/haowjy/creative-writing-skills --skill prose-writing

简介

prose-writing 用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写,适合让 Agent 提炼结构、补齐章节、统一术语或把零散材料整理成可读文档。

  • 适用于技术文档编写、内容稿件优化和文案结构重组等场景。
  • 支持对现有文本进行格式调整、逻辑梳理和语言润色,保留原始事实。
  • 使用时应避免过度营销或夸大能力,涉及对外文案时需控制语气。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

Prose Writing

Draft narrative fiction that serves the story's needs. This skill covers craft technique — the how of prose. Project-specific voice, character speech patterns, and formatting conventions come from style files. Read those before writing.

Two Modes

Interactive

Back-and-forth with the author. They describe what they want, you draft, they react, you revise. The conversation IS the feedback loop — don't produce a "final" draft on the first pass. Offer options when the direction is ambiguous. Ask about intent when a scene could go multiple ways.

Autonomous

You receive a brief (scene outline, beat list, emotional arc, relevant context) and produce a draft. No author in the loop — the draft goes to critics next, not to a human. This means:

  • Follow the brief's direction without second-guessing it. The orchestrator already confirmed intent.
  • Make concrete choices rather than hedging. A critic can flag a bad choice; they can't critique a hedge.
  • Write the full scene, not a summary or outline. The output is prose, not planning.
  • Mark places where you made significant interpretive choices (e.g., <!-- CHOICE: interpreted "tense conversation" as passive-aggressive rather than openly hostile -->) so critics and the orchestrator can evaluate your calls.

Scene Entry

Open scenes in the middle of something happening. The reader should orient through action and context, not through setup narration. A character already mid-task, mid-conversation, or mid-thought gives the reader something to track immediately.

Avoid:

  • Weather/setting paragraphs before anything happens
  • Character waking up or traveling to the scene location
  • "The room was..." description blocks
  • Recap of how we got here

The exception is when the setting itself IS the story beat — a character seeing a destroyed city for the first time, or arriving somewhere that changes everything. Then the setting description carries narrative weight.

Scene Purpose

Every scene needs to change something — character knowledge, relationships, power dynamics, stakes, emotional state. If nothing is different at the end than at the beginning, the scene is a candidate for cutting.

Before writing, identify what changes. In interactive mode, ask the author. In autonomous mode, extract it from the brief. If the brief doesn't specify what changes, flag it: <!-- UNCLEAR: brief doesn't specify what changes in this scene -->.

Point of View

Stay in the established POV. First person means the narrator can only report what they perceive, think, and feel — not what other characters are thinking. Third limited means access to one character's interiority per scene.

POV violations are the most common craft error in AI-generated fiction. Watch for:

  • Reporting emotions of non-POV characters as fact ("She felt angry") instead of through observable behavior ("Her jaw tightened")
  • The narrator knowing things they shouldn't (what happened in a room they weren't in)
  • Head-hopping between characters within a scene

Show Through Action

Demonstrate character states through behavior, dialogue, and physical response rather than narrating them directly.

Telling: "He was nervous about the meeting." Showing: "He checked his watch for the third time, then straightened a tie that was already straight."

This doesn't mean never tell. Summary narration is appropriate for:

  • Transitions between important moments
  • Routine actions that don't carry emotional weight
  • Time compression ("The next three weeks passed in a blur of training")

The heuristic: tell for logistics, show for moments that matter.

Dialogue

Dialogue should do at least two things simultaneously: advance the plot AND reveal character, or reveal character AND build tension, or build tension AND seed information. Single-purpose dialogue ("As you know, the reactor is on the third floor") feels flat because real conversation is never purely transactional.

Subtext. Characters rarely say exactly what they mean. They deflect, understate, change the subject, answer a different question than the one asked. The gap between what's said and what's meant is where characterization lives.

Voice differentiation. Each character should sound distinct enough that you could identify the speaker without dialogue tags. This comes from vocabulary, sentence structure, speech patterns, and what they choose to talk about — not from accent spelling or verbal tics bolted on.

Action beats over dialogue tags. "Said" is invisible; use it freely. Avoid creative tags ("he exclaimed," "she retorted") — they draw attention to the tag instead of the dialogue. Use action beats to show how something is said: "She set the cup down carefully. 'That's not what I meant.'"

Pacing

Vary sentence length and structure to control reading speed. Short sentences speed things up — good for action, tension, shock. Longer sentences slow the reader down — good for reflection, description, complex emotion.

Scene-level pacing: alternate between high-tension and lower-tension beats. Sustained intensity becomes numbing. The quiet moment after the crisis is what gives the crisis weight.

Chapter-level pacing: end on forward momentum. Not necessarily a cliffhanger — an unanswered question, a new complication, an emotional shift. Give the reader a reason to continue.

Sensory Detail

Ground scenes in specific sensory details rather than generic description. "The forest" is nowhere. "Pine needles crunching underfoot, the smell of wet bark" is a place.

Use the POV character's sensory priorities. A chef notices smells. A musician notices sounds. A soldier notices exits. What a character pays attention to reveals who they are.

Avoid cataloging all five senses mechanically. One or two vivid, specific details do more work than a comprehensive sensory inventory.

Interiority

The POV character's inner life — thoughts, reactions, memories, associations — is what distinguishes prose fiction from screenplay. Don't neglect it.

But interiority should be specific and revealing, not generic emotional labeling. "She felt sad" is a label. "She kept reaching for her phone to text him before remembering" is interiority that shows the reader something about grief.

In high-action scenes, interiority contracts to fragments — quick reactions, snap judgments. In reflective scenes, it expands. Match the depth of inner experience to the pace of the scene.

Revision Signals

When reviewing your own draft (or receiving critique), watch for these patterns:

  • Redundant beats. Saying the same thing through action, dialogue, AND narration. Pick the strongest delivery and cut the others.
  • Throat-clearing. Opening paragraphs that circle the point before landing on it. Cut to where the real scene starts.
  • Emotional choreography. Characters cycling through emotions in predictable patterns (shock → denial → anger → acceptance) without the specific, messy reality of how people actually process things.
  • Placeholder prose. Competent but generic sentences that could appear in any story. If a sentence could be swapped into a different scene without anyone noticing, it's not doing enough work.

Resources

Style files passed alongside this skill contain project-specific voice guidance — character speech patterns, tonal registers, formatting conventions, scene-type techniques. Read them before drafting. They override general craft advice when they're more specific.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35%
按下载量换算255

Claude

27.13%
按下载量换算198

Cursor

20.55%
按下载量换算150

Gemini CLI

10.27%
按下载量换算75

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills