wallstreet-explainer skill
wallstreet-explainer is a video skill that ties spoken lines to scene templates and supports custom code for individual shots.
- 566 ★ on the repository
- Last commit
Extended access
Why this level
- The agent downloads media and writes video, voice and project files.
- Scripts use a browser and local media tools; the skill stages include approval steps.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/wallstreet-explainerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent downloads media and writes video, voice and project files.
- Scripts use a browser and local media tools; the skill stages include approval steps.
This is third-party code. Review the repository files before installing.
What it does
wallstreet-explainer is a Claude Code skill that guides an agent from a finance or technology topic to a completed video. The agent finds sources, writes a script, assembles footage and submits intermediate results for user approval. Material origins are recorded. In template mode, scenes come from a library and are tied to spoken phrases. The agent can code an individual shot when a template does not fit. Included scripts generate temporary speech, build timing, render frames and mix audio.
Who it is for. Finance and technology video authors who review the script, sources and editing at successive stages.
Good fit when
- An author explains money flows between companies with numbers and charts.
- An editor turns a report and official footage into a short sourced video.
- After script approval, a team replaces temporary speech with final voice recordings.
- A single shot needs custom code while the rest of the project uses templates.
Not a fit when
- The production uses an on-camera presenter.
- A producer expects a finished release without reviewing scripts or intermediate shots.
- The author lacks permission to use the selected third-party footage in the published video.
Example request
Prepare a topic and script for a one-minute video about AI spending, show primary sources and wait for approval before voice production.Limitations
Production requires Python, Node.js, Playwright with Chromium and ffmpeg; optional processing steps have further dependencies. Fonts are not bundled and must match the names referenced by `kit/kit.css`. Third-party footage, screenshots and people’s images are outside the project’s MIT license. The author has not yet tested whether template mode reliably works with weaker models, so that remains an intended benefit rather than a demonstrated result.
How to disable. Remove the skill directory from Claude Code’s skill paths and restart the agent. Stop render processes separately; saved projects, footage and videos remain in their folders.
FAQ
How do I add the skill to Claude Code?
Give the agent https://github.com/HeyChengC/wallstreet-explainer to import. The manual Claude Code method clones into `~/.claude/skills/wallstreet-explainer` and installs Python and Node dependencies.
How do I check the shot plan before rendering?
From the skill directory run `node scripts/render.mjs --dir <片子> --check`, replacing `<片子>` with your project directory. It reports scene errors and warnings without rendering a video; resolve them before producing frames.
How do I replace temporary speech?
Save each line as `voice/<句子id>.final.wav`, replacing the placeholder with its ID. Then use the README example `python scripts/temp_voice.py film/script.json --asr` to realign speech timing.
How do I inspect shots before rendering the video?
The skill’s `render.mjs --sheet` assembles three frames per shot into `out/sheet.png`. Request that sheet after plan validation and review it before a full render.
Can templates and custom code be mixed?
Yes. The agent can use `scene:'custom'` and write code for one shot while keeping other scenes templated. Retain `says`, the shot’s intended message, and review that shot before rendering the full video.
Discussion
No questions or reviews yet.
Related
SkillsA Claude Code and Codex skill that turns a topic into a narrated motion-graphics explainer video on Remotion with subtitles and chapters
The video-explainer skill turns a script into a vertical 9:16 explainer: narration, captions, Remotion and a pre-render review
Claude Code skill that turns one topic into a 60-second Vox-style explainer: script, keyframes, animation, voiceover and music.
Skills for faceless video channels: one topic becomes a video with voiceover, subtitles, covers and per-platform copy.
A skill and CLI: the agent writes a script and scenes, while the voice, word timing and MP4 render run on your computer with no API keys
A video editing skill for Claude Code and Codex: the agent cuts pauses and filler words, grades color, burns subtitles and adds animated overlays
README in short
`SKILL.md` links to script, motion and sourcing rules. Tools cover video search, news screenshots, frame rendering and audio mixing. Template and hand-coded shots can coexist in a project.
README badge
Original SKILL.md
--- name: wallstreet-explainer description: 华尔街式的财经/科技解读短片（1–3 分钟、16:9、没有主持人出镜）：用真实的人、真实的新闻和官方画面作证据；当事人的实拍抠像替代主持人（人站在巨字前面、人身后换背景、横推换人）；图形（照片圆盘、金色大数字、柱子和图表）只用来演钱和关系，底是一张细斜纹深蓝织物，镜头在有前后层的 2.5D 世界里走，每一镜都接住上一镜往下推。两档做法：模板档（kit：一条片 = 一张镜头表，选场景、填内容，镜头和动效由模板管，render.mjs --check 用文字自检，目标是弱一些的模型也能稳定出片）和手写档（强模型逐镜手写）。配一套工具：找片下片抽帧、新闻原文截图、GPU 逐帧抠人、临时配音和时间线、逐帧渲染、镜头运动自检、混音。用户说"wallstreet-explainer""做一条财经/科技解读""把一笔账讲清楚""巨头砸钱/估值/营收/资金循环这类题目"时使用；有真人口播出镜的片子不用这个。 --- # 华尔街式解读片（wallstreet-explainer） **讲一笔大账：真的人、真的新闻、真的画面作证据，图形舞台把数字和关系演出来；每一镜都接住上一镜的东西往下走，观众始终知道“现在在看什么、为什么看这个”。** **动的是镜头和世界，不是元素自己演：镜头在有前后层的舞台里推、拉、摇，去找正在说的那个东西；东西在原地换状态。每一下动作都对着那句话的意思。** ## 这份 skill 是什么 - **是**：一套设计思路（每一镜放什么、为什么、凭什么接下一镜）、一套动效原理（从三条同类参考片逐帧量出来：为什么它不像 PPT、不像 AI，手法清单和量化目标）、一套风格语言（舞台、巨字、抠像、圆盘、数字、图表、原文高亮的具体做法，外加一份开头钩子菜单），以及一套工具。 - **两档做法，画风是同一个：** - **模板档（默认，先用这个）**：`kit/` 里 33 个场景（大数字、照片圆盘、新闻桌面、原文刷黄、人站在巨字前面、柱子、图表、地图……）。一条片 = 一张镜头表：每一镜选一个场景，填这一镜的字、数、图和“念到哪个字”。镜头怎么走、东西怎么出现、前后层、运动模糊、字号颜色，都由模板管。这样设计，是想让弱一些的模型也能稳定出片；这一点还没实测过。 - **手写档（强模型，或者模板对不上的时候）**：底座 `film_base.js` + 手法库 `moves.js` + 2.5D 小工具 `stage25d.js`，逐镜手写。只有一镜对不上时，在镜头表里写 `scene:'custom'`，手写那一镜就行。 - **模板是词汇，不是镜头顺序**。每条片的画面都要从它自己的文案和素材里现想：每一镜先写 `says`（这一镜让观众多懂了什么），再找意思对得上的场景；只是差不多就换一个，或者手写。`examples/` 里的两条片记录的是片子怎么想出来、在哪里摔过跤，不要照搬它们的镜头。 - **画风的基准是成片，不是模板**：《AI到底有多烧钱》的每秒一帧（`examples/ai-burn-money/frames_1fps.jpg`）和 28 秒的《谷歌给苹果的两百亿》（`examples/google-apple-20b/`，真素材手写）。模板的默认样子就是照它们做的；做出来的片子要能和它们并排放而不突兀。 - **手写档能复用的有两样，都是为了质量稳定**：底座（`scripts/film_base.js`，每条片都一样的那层）和手法库（`scripts/moves.js`，磨过、逐帧查过的动作）。手法是词汇，不是镜头：先想这一镜要说什么，再找意思最对得上的手法；只是差不多就手写。不要为了用上手法去改画面，也不要为了出片去凑手法（选法见 `references/motion.md` 的手法库）。 - **慢推、数字滚动、实拍上压字，参考片都在用**。廉价的不是这些手法，是做法。判断用三句话：删掉它观众少懂了什么？是镜头在动、东西在换状态，还是元素自己在演入场？这一层是真的、具体的吗？ 动手前先读 `references/thinking.md`、`references/motion.md` 和 `references/rules.md`；做模板档再读 `kit/README.md`。其他几份按需要读。 ## 五条硬规矩（完整的在 `references/rules.md`，带编号、能检查） 1. **真东西是主角**：当事人实拍、新闻原文截图、官方发布会/财报/广告画面。图形只用来演钱和关系。示例片里的占位图（`kit/assets/ph/`）在正式片里一张都不能留。 2. **底只有一张细斜纹深蓝织物**。不要满屏 3D 物体，不要紫、青这类整屏底色，不要大团彩色光，不要浮尘。品牌色只在当事人身后出现；酒红底一条片最多一段（算总账那段）。 3. **每一镜写 `says`**，一镜一个主角；东西朴素地出现，镜头去找它。 4. **时刻挂在字上**（`'@b05:英伟达'`），不写死秒数；开头 0.5 秒内开口，第一帧就有东西。 5. **交片前**：`render.mjs --check` 没有错误；`--sheet` 自己看过；720p 抽帧和 `frames_1fps.jpg` 并排比过；关键动作看过连续帧。 ## 流程：每一站停下来给用户看 审美由用户把关。不要一口气出成片；每站给出具体的东西，等他点头再往下走。 | 站 | 做什么 | 给用户看什么 | |---|---|---| | 0 选题 | 题目要同时满足“现在热”和“看完能带走点什么”；事实先查清、留链接 | 两三个题目，每个附一句钩子 | | 1 文案 | 口语、第一人称，句句往下接；数字写成读法；插原声的地方先写铺垫句（`references/audio.md`） | 全文。点头之前不做正式配音 | | 2 导演表 | 见下面“导演表”一节。模板档在“手法”一列写场景名；每一镜的素材从哪来也写上 | 要点：开头怎么抓人、每处怎么接、要不要加铺垫句、哪几处用当事人抠像 |