内容写作
- 作者仓库星标 834
- 作者更新于 实时读取
- 作者仓库 boop-agent
- 领域
- 效率工具
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @raroque · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: youtube-script-writer
description: Write a tight, retention-focused YouTube script from a topic, outline, or research brief. Use wh…
category: 效率工具
runtime: 无特殊运行时
---
# youtube-script-writer 输出预览
## PART A: 任务判断
- 适用问题:日常重复事务、资料整理、邮件或工作流提效。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“When to use / When NOT to use / Inputs to gather first”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于日常重复事务、资料整理、邮件或工作流提效,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“When to use / When NOT to use / Inputs to gather first”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文没有稳定的斜杠命令要求。安装验证后通常全局生效,直接在对话里点名这个 Skill 并描述任务即可。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件。
先用一个小任务确认它会围绕“When to use / When NOT to use / Inputs to gather first”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: youtube-script-writer
description: Write a tight, retention-focused YouTube script from a topic, outline, or research brief. Use wh…
category: 效率工具
source: raroque/boop-agent
---
# youtube-script-writer
## 什么时候使用
- 把日常协作方向的常用动作沉淀成 Agent 可调用的技能 适合处理重复事务、信息整理、邮件和日常工作流提效,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代的步骤…
- 面向日常重复事务、资料整理、邮件或工作流提效,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「When to use / When NOT to use / Inputs to gather first」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "youtube-script-writer" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> When to use / When NOT to use / Inputs to gather first
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} YouTube Script Writer
Turn a topic, outline, or research dump into a YouTube script that holds attention from second 0 through the CTA.
When to use
- "Write a YouTube script about X"
- "Turn this research into a video"
- "Draft an intro / hook for a video on Y"
- "Plan out a 5-minute video about Z"
When NOT to use
- Long-form blog posts or articles (different pacing rules — don't force them into this template)
- Podcast outlines (conversational, not scripted)
- Shorts under 60s (different structure — lead with the payoff, no setup)
Inputs to gather first
Before writing, confirm (ask if missing):
- Topic / angle — what is the one idea the viewer walks away with?
- Target length — under 3 min, 3–8 min, or 10+ min? Pacing changes.
- Audience — beginners, experts, or mixed? Sets vocabulary floor.
- Research or source material — any facts, stats, quotes, URLs to include?
- Tone — casual/educational, docu-style, chaotic/entertainment?
If the user didn't supply research and the topic needs facts, use WebSearch / WebFetch BEFORE writing. Never invent stats or quotes.
Script structure (standard 5–8 min video)
Write the script in these labeled sections. Don't skip any.
1. Hook — first 5 seconds
- One sentence that promises a payoff or opens a loop.
- No "hey guys welcome back." No throat-clearing.
- Patterns that work:
- Curiosity gap — "Everyone gets X wrong. Here's what actually works."
- Bold claim — "You can do X in 10 minutes. I'll prove it."
- Pattern interrupt — visual or verbal surprise.
- End with an implicit "keep watching to find out."
2. Setup — seconds 5–20
- Context: why does this matter NOW, for THIS viewer?
- Credibility beat (one sentence max): who you are / why listen.
- Restate the promise in sharper form. No long preamble.
3. Body — main content
- Break into 2–4 clearly-labeled sections/chapters.
- Retention rule: every 30–45 seconds, re-hook. Tease what's coming ("but here's the twist…"), change visual pattern, or drop a memorable line.
- One idea per section. Don't stack.
- Concrete > abstract. Examples, numbers, screenshots, demos.
- If you have research: cite specifically (brand, study, date). No vague "studies show."
4. Payoff — deliver on the hook
- Close the loop opened in the hook.
- Summarize the one takeaway in one sentence. Viewer should be able to repeat it.
5. CTA — final 15 seconds
- ONE ask. Not three. Pick: subscribe, comment prompt, link in description, or watch-next video.
- Make it natural, tied to the content. "If this saved you X, subscribe" beats "smash that like button."
- End on a strong visual beat, not a fade.
Format the output
Deliver the script as:
TITLE: <working title, clickable but honest>
THUMBNAIL IDEA: <1 sentence, what's on screen>
---
[0:00] HOOK
<exact words to say>
[0:05] SETUP
<exact words>
[0:20] SECTION 1 — <name>
<exact words>
[VISUAL: brief note on B-roll / cut / graphic]
[...continue...]
[X:XX] CTA
<exact words>
Rules for the script body:
- Write the spoken words, not stage directions, as the main text.
- Put visual/cut notes in
[VISUAL: …]tags inline, indented under the spoken line they belong to. - Use short sentences. Read it aloud in your head — if a sentence is hard to say, rewrite it.
- Contractions always (don't → that's how people talk).
- No jargon without a one-line definition.
After the script
Close with:
- Estimated runtime — rough estimate based on ~150 wpm spoken pace.
- Open questions — anything you had to guess at (e.g. target audience assumed to be X; pick a different tone if wrong).
- Variants to consider — one alt hook and one alt CTA, one line each. Gives the user pivot options.
Retention checklist (self-review before delivering)
- Hook is under 10 seconds and makes a clear promise
- Setup ends in under 30 seconds total
- A re-hook or pattern break every 30–45 seconds of body
- Every claim has a source or is clearly framed as opinion
- Exactly one CTA
- Body reads aloud cleanly — no tongue-twisters, no buried verbs
- Sources section follows (if WebSearch/WebFetch was used)
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核