Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add chenyuxiaojin/video-agent-skills --skill video-agent-writergit clone --depth 1 https://github.com/chenyuxiaojin/video-agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/chenyuxiaojin/video-agent-skills/video-agent-writer)<a href="https://agentmods.dev/skills/chenyuxiaojin/video-agent-skills/video-agent-writer"><img src="https://agentmods.dev/badge/skills/chenyuxiaojin/video-agent-skills/video-agent-writer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/chenyuxiaojin/video-agent-skills/video-agent-writer"><img src="https://agentmods.dev/badge/skills/chenyuxiaojin/video-agent-skills/video-agent-writer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00107 | $0.02033 |
| Opus 5 | $0.00053 | $0.01017 |
| Sonnet 5 | $0.00021 | $0.00407 |
| Haiku 4.5 | $0.00011 | $0.00203 |
Grade A, and why
video-agent-writer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.
video-agent-writer（编剧）
职责边界
编剧只负责文字内容。具体分工：
- ✅ 逐字稿撰写（叙事、论点、案例、金句）
- ✅ 段落划分
- ✅ 节奏控制（信息密度、情绪起伏的文字节奏）
- ✅ 出处标注
- ❌ 详细画面设计（分镜师负责）
- ❌ 素材搜索（美术负责）
- ❌ 时间轴（剪辑师负责）
输入 → 输出
- 输入：
outline.md - 输出：
script.md
逐字稿格式
# 视频标题
> 逐字稿 | 目标时长：约 X 分钟 | 开头模式：反常识/故事/问题
---
## 开头（0:00-0:30）
你有没有想过一个问题：为什么你学了那么多时间管理方法……
---
## 案例一：小标题 — 从XX角度看（0:30-2:30）
我们先来拆解一个陷阱……
Cal Newport 在《Deep Work》这本书里提出了一个概念……
【出处：Cal Newport,《Deep Work》第一章】
加州大学的 Gloria Mark 教授做过一项研究……
【出处：Gloria Mark, UC Irvine 注意力研究】
关键格式规则：
- 提到人名时保持全名，分镜师会据此安排人物照片
- 提到书籍/论文时标明完整名称，分镜师会安排封面画面
- 数据和统计单独成句，方便分镜师设计数据动效
- 出处用
【出处：...】标注，独立一行
执行步骤
步骤 1：读取大纲，识别主题类型，加载模板
- 读取
outline.md - 识别大纲顶部的
主题类型字段（如cognitive、tech-sci、review等） - 读取
templates/<主题类型>.md模板文件，获取该类型的开头模式、结构模板、案例组合和过渡句 - 如果大纲未标注主题类型，根据内容自行判断并告知用户
- 如果对应模板不存在，使用通用框架（步骤 3 的案例结构）并告知用户
- 确定核心主题、目标时长、论点数量
- 参考字数与时长对应表：
| 时长 | 字数（280字/分钟） | 支撑案例数 |
|---|---|---|
| 5 分钟 | ~1400 字 | 2 个 |
| 8 分钟 | ~2240 字 | 3-4 个 |
| 10 分钟 | ~2800 字 | 4-5 个 |
| 15 分钟 | ~4200 字 | 5-6 个 |
步骤 2：选择开头模式
参考模板文件中的「开头模式表」，根据主题选择合适的开头方式。
通用原则：开头 30 秒内必须有一个让观众停下来的钩子。三种基本模式：
- 反常识：打破常见认知，制造信息差
- 故事：用具体场景带入，引发共鸣
- 问题：直击痛点，让观众想知道答案
步骤 3：建立结构 + 案例/论证框架
参考模板文件中的「结构模板」和「案例组合」，按主题类型选择合适的结构。
通用框架（所有类型适用）：
每个案例/论证层按以下结构展开：
- 引入：一句话衔接上文，引出这个部分的角度
- 故事/数据：具体的人、事、数据（时间、地点、人物、经过、结果）
- 出处标注：独立一行
- 回扣/小结：一句话收束，衔接下一部分
关键原则：
- 各部分的角度要有差异，避免重复论证同一个面
- 部分之间用过渡句连接（参考模板中的过渡句示例）
- 具体的结构模板、案例组合方式、过渡句风格见对应模板文件
- 破立结合：每个段落先指出问题 → 紧接给出解法/启发/获得感。禁止连续 2 段以上纯批评或纯问题陈述。自检：每段结束时，观众能带走什么？
叙事视角
- 默认第一人称口播：用"我""我发现""我有个朋友"，不要用"某某说""据报道"等第三方旁白
- 引用他人观点时，融入自己的表达：❌「Sam Altman 说 AGI 即将到来」→ ✅「我看到 Sam Altman 有个判断让我特别在意——他说 AGI 可能比我们想的更近」
- 让观众感觉是"一个真人在跟我聊"，不是"一个旁白在念稿"
素材富裕模式
当整理稿（outline.md）已经非常丰富（≥ 原素材 60%、含原文引用和金句索引）时：
- 你的工作是语言转换（书面体→口播体），不是从零创作
- 保留原文力量：整理稿中标注的原文引用和金句，尽量保留原始表达的力量，只做口语化适配
- 不要过度改写：如果原文已经说得好，直接用，加上口播过渡即可
- 编剧指引中标注的情绪基调和建议时长，直接遵循
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 178 lines · 107 tokens per session scan A 48a8d640e8fa
video-agent-writer is a skill published in the GitHub repository chenyuxiaojin/video-agent-skills (9 stars, last pushed 3mo ago), licensed MIT. It adds 107 tokens to every session and 2,033 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
suggest-sfx
The SFX pass. Analyze a short's beats + narration and propose tasteful sound effects synced to them, drawing from (and growing) a shared, reusable SFX library, then render an SFX-mixed audition preview. Use when the user wants to "add SFX / sound effects", "suggest sfx", "score the transitions", "sound-design this…
make-ai-short
Build a GENERATIVE-pixels vertical short (1080×1920, 35-40s) end-to-end — a recurring locked character animated by a fal video model (Seedance 1.5 Pro default) under a philosophical/story VO, composited in Remotion with word-synced captions, SFX audition, seamless frame-0==last-frame loop. Use when the user wants to…
make-short
Build a fully-synthetic vertical short (1080×1920, 40s) end-to-end from a topic — script + beats.json, linked TSX beats over a persistent canvas, phone-scale QA, render, ElevenLabs voice with exact word-synced captions, SFX audition, optional music bed. Use when the user wants to "make a short", "create a shorts…
make-vox
Build a Vox-style LAYERED-COLLAGE vertical short end-to-end — script + scene dissection into image layers, layer production (AI images via genimage.py + cutout.py, HTML→PNG, SVG-in-TSX), choreographed TSX assembly on the collage kit (CollageBoard camera, parallax, cutouts, chips, routes), frame QA, render, then…
talking-head-edit
Edit interviews, lessons, commentary, podcasts, and direct-to-camera recordings built around speech. Use it to remove dead air, filler words, false starts, repeated lines, and discarded retakes while preserving natural delivery and keeping captions aligned. When the user asks for visual enhancement, also use it for…
montage-edit
Discover the governing emotion and visual motifs in a body of footage, then compose a montage whose image, rhythm, sound, and text form one intentional experience.