open-video-harness git:20260928.98dcd0fB

Immutable. This exact content never changes and is served at /api/v1/blob/c03280537329f29a.

---
name: open-video-harness
description: Make videos with code (video-as-code / code2video) through the OpenVideoHarness workbench. Covers 3Blue1Brown-style math and science explainers, knowledge shorts (vertical or horizontal; 抖音, B站, 小红书, Shorts), product launch films and promos, lyric videos and music videos, data stories, paper explainers, hand-drawn or watercolour shorts, and meme or fast-cut edits. Also covers voiceover (Chinese or English), bilingual captions, code-composed music, sound effects, and reverse-engineering a reference video. Use it whenever the user asks to make, animate, render or remake a video, 做视频, 做动画, 做科普视频, 做宣传片, 做 MV, 讲解视频, or 片头, even if they don't name the harness.
---

# OpenVideoHarness

Harness path (set by install.sh): ~/OpenVideoHarness

This skill is a thin pointer. All workflows, rules and tools live in the OpenVideoHarness repository.

## 1. Locate the harness

Use the first of these that exists and contains `bin/vh`:

1. `$OVH_DIR`
2. the path on the "Harness path" line above
3. `~/OpenVideoHarness`
4. the current working directory

If none exists, ask the user for permission, then install it:

```bash
curl -fsSL https://raw.githubusercontent.com/ZLHad/OpenVideoHarness/main/install.sh | bash
```

## 2. Follow the harness, not this file

Read `<harness>/CLAUDE.md` (Codex: `<harness>/AGENTS.md`, which is identical) and follow it exactly:

- route the request to one of the 8 `video-types/` docs;
- create the project with `<harness>/bin/vh new <type> <slug>`;
- respect the **three human review gates** (outline → storyboard with a keyframe preview → first draft); stop at each gate and wait for the user;
- follow the hard rules: every frame is a pure function of t; when there is sound, audio sets the timing; storyboard before code; self-review every scene; fact discipline;
- keep work inside `<harness>/projects/`, or in a directory the user names.

Audio tools:
- `bin/vh tts` — voiceover, Chinese or English, local Qwen3-TTS by default;
- `bin/vh captions` — zh/en/bilingual subtitles;
- `bin/vh music` — code-composed soundtrack;
- `bin/vh sfx` — sound effects;
- `bin/vh mix` and `bin/vh mux` — mixing and muxing.

Run `bin/vh doctor` if anything looks missing.