Agent skill

Claude Real Video

by HUANGCHIHHUNGLeo in HUANGCHIHHUNGLeo/claude-real-video

幫 Leo 看影片。當 Leo 丟影片連結(YouTube/IG/TikTok)或本機影片檔,要摘要、分析、拆解對標時使用——Claude 不能直接吃影片,先用這個工具抽關鍵幀+逐字稿+運鏡節奏+聲音/語氣/手勢時間軸再讀。

MITAuto-check passedMedia & Creative

Install Claude Real Video

skills CLI
$ npx skills add HUANGCHIHHUNGLeo/claude-real-video --skill claude-real-video -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install HUANGCHIHHUNGLeo/claude-real-video claude-real-video --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/HUANGCHIHHUNGLeo/claude-real-video.git skills-src && mkdir -p .claude/skills && cp -r skills-src/personal-skill-backup .claude/skills/claude-real-video && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
claude-real-video
GitHub stars
2.2k
Token cost
~559 tokens
SKILL.md length
128 words
Files
1
Skills in repo
3
Repo updated
First seen
Licence
MIT

At a glance

幫 Leo 看影片。當 Leo 丟影片連結(YouTube/IG/TikTok)或本機影片檔,要摘要、分析、拆解對標時使用——Claude 不能直接吃影片,先用這個工具抽關鍵幀+逐字稿+運鏡節奏+聲音/語氣/手勢時間軸再讀。

  • Works in 3 steps: MANIFEST.txt — 開頭有 lens… → grids/ 九宮格連續幀;要看細節才開 frames/*.jpg(讀前縮… → perception.json — 完整感知資料(被門檻藏掉的低信心事件在這)
  • Media & Creative work in your project
  • SKILL.md covers 什麼時候用, 指令怎麼選(0.5.0), 讀輸出的順序 and ⚠️ 鐵則(2026-07-05 Leo QA 定的), plus 1 more section
  • Calls python3

What it does

Claude Real Video is an agent skill from HUANGCHIHHUNGLeo/claude-real-video. 幫 Leo 看影片。當 Leo 丟影片連結(YouTube/IG/TikTok)或本機影片檔,要摘要、分析、拆解對標時使用——Claude 不能直接吃影片,先用這個工具抽關鍵幀+逐字稿+運鏡節奏+聲音/語氣/手勢時間軸再讀。

Its SKILL.md is about 560 tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Media & Creative. It works with TikTok, YouTube, FFmpeg and Python. The repository describes itself as: Let Claude (or any LLM) actually watch a video — scene-aware, deduplicated frames + transcript, from a URL or local file. Runs locally, MIT. The licence is MIT.

When your agent uses it

  • Media & Creative work in your project

Example prompts

  • “/claude-real-video”

Requirements

  • Python 3

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

  1. MANIFEST.txt — 開頭有 lens 指令區塊照著做;含幀清單、motion 區塊、perception timeline(聲音/語氣/情緒/手勢/畫面事件,各帶秒數與信心分數)、逐字稿
  2. grids/ 九宮格連續幀;要看細節才開 frames/*.jpg(讀前縮 1500px 內)
  3. perception.json — 完整感知資料(被門檻藏掉的低信心事件在這)

What it can do on your machine

Read from SKILL.md and the folder at commit ab4b9b2. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python3

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Claude Real Video loads about 559 tokens when it runs. Until then it costs about 33 tokens; SKILL.md has 128 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~33
When it runs · the whole SKILL.md, loaded when a task matches
~559

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from HUANGCHIHHUNGLeo/claude-real-video at commit ab4b9b2, republished under its MIT licence (© HUANGCHIHHUNGLeo). 128 words, ~559 tokens.

Download SKILL.mdSave it as .claude/skills/claude-real-video/SKILL.md (or your agent's skills folder).
name
claude-real-video
description
幫 Leo 看影片。當 Leo 丟影片連結(YouTube/IG/TikTok)或本機影片檔,要摘要、分析、拆解對標時使用——Claude 不能直接吃影片,先用這個工具抽關鍵幀+逐字稿+運鏡節奏+聲音/語氣/手勢時間軸再讀。

claude-real-video / crv Pro — 幫 Leo 看影片

Leo 本機裝的是 Pro 版 0.5.0(~/Projects/crv-pro/.venv/bin/crv-pro,PATH 有的環境直接 crv-pro)。

什麼時候用

Leo 給影片(網址或檔案路徑)問內容、要摘要、拆對標 reel、問「他怎麼講/怎麼拍/觀眾為什麼買單」時。

指令怎麼選(0.5.0)

Leo 的話聽起來像指令
「這在講什麼」「幫我摘要」「他說的是真的嗎」crv-pro "<src>" -o out --mode watch
「怎麼做的」「拆解」「為什麼會紅」「能學嗎」crv-pro "<src>" -o out --mode creator
「完整分析」或意圖不明crv-pro "<src>" -o out --mode full

Leo 有具體問題就加 --why "<他的原話>"。

  • --senses = 五感全開:聲音事件、語氣曲線、情緒、手勢/表情、畫面標籤(首次跑下載模型 ~2.7GB,之後 86 秒影片約 40 秒)
  • 只要單一感官:--prosody / --audio-events / --emotion / --gesture / --scene-labels
  • --lens content|creator|both(預設 both)決定 MANIFEST 教讀的 AI 往哪個方向分析
  • 長影片加 --max-frames 60;無語音影片 --no-transcribe
  • IG/TikTok 這類要登入的平台:先把 cookie JSON 轉 Netscape 格式再走 --cookies(2026-07-05 實測可直接抓 IG reel,這才是正路):
    bash
    python3 -c "import json; ck=json.load(open('/Users/leo/Projects/auto-post/browse-cookies/kanisleo328-threads.json')); cs=ck if isinstance(ck,list) else ck['cookies']; print('# Netscape HTTP Cookie File'); [print('\t'.join([c['domain'],'TRUE' if c['domain'].startswith('.') else 'FALSE',c.get('path','/'),'TRUE',str(int(c.get('expirationDate',2e9))),c['name'],c['value']])) for c in cs if 'instagram' in c.get('domain','')]" > /tmp/ig-cookies.txt
    crv-pro "<reel url>" --cookies /tmp/ig-cookies.txt --mode full
    這條也失敗才退 [[reference_ig_post_full_read]] 的 IG 內部 API 手動下載

讀輸出的順序

  1. MANIFEST.txt — 開頭有 lens 指令區塊照著做;含幀清單、motion 區塊、perception timeline(聲音/語氣/情緒/手勢/畫面事件,各帶秒數與信心分數)、逐字稿
  2. grids/ 九宮格連續幀;要看細節才開 frames/*.jpg(讀前縮 1500px 內)
  3. perception.json — 完整感知資料(被門檻藏掉的低信心事件在這)

⚠️ 鐵則(2026-07-05 Leo QA 定的)

  • 逐字稿必須從頭讀到尾才准出報告(2026-07-16 Leo 抓到抽樣讀導致講錯出處+漏掉全片最有哏的三個細節):transcript.txt 全文讀完再寫內容拆解,抽樣只能用來對秒數,不能替代理解

  • 感知標籤是線索不是定論:引用前對照幀和逐字稿交叉驗證(MANIFEST 開頭也這樣教);情緒模型會把「正常有力的語氣」誤讀成 angry——非中性情緒信心 <0.75 已被藏,露出的也要驗

  • 拆對標/做復刻前必看 motion 節奏表(cuts/min、中位鏡頭長、運鏡分佈),別憑感覺(詳 [[reference_reel_clone_pipeline]])

  • 回答引用具體秒數和數據,不籠統

備註

  • 全程本機跑,工具不上傳任何東西;--kb <目錄> 存進知識庫
  • 暫存的下載影片/抽幀讀完要清(/tmp 下的)

© HUANGCHIHHUNGLeo, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in personal-skill-backup of HUANGCHIHHUNGLeo/claude-real-video.

Open the folder on GitHubat commit ab4b9b2

Compare with similar skills

Claude Real Video next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Claude Real Video compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Claude Real Video this skillHUANGCHIHHUNGLeo/claude-real-video2.2k—~559Automated safety check: PassMIT
Ffmpeg Skillkajisho5/ffmpeg-skill1.9k—~7.4kAutomated safety check: PassMIT
Watchmathiaschu/watch142—~4kAutomated safety check: WarnMIT
Youtube Clipperop7418/Youtube-clipper-skill2.2k—~1.6kAutomated safety check: NotesMIT
Video Transcribewendy7756/AI-Video-Transcriber3.3k—~937Automated safety check: NotesApache-2.0
Explainroovincentsch/explainroo515—~439Automated safety check: PassMIT

Similar skills

  • Ffmpeg Skill

    kajisho5/ffmpeg-skill

    Edit video and audio with local FFmpeg from natural-language requests: cut, trim, join, resize/reframe (9:16, 1:1), speed change, captions and subtitles (SRT/ASS, animated, karaoke), logos and text…

    1.9k GitHub stars~7.4k tokensUpdated yesterday
    Media & CreativeAuto-check passed
  • Watch

    mathiaschu/watch

    Watch a video from YouTube, Instagram, X/Twitter, Vimeo, TikTok or any of ~1800 yt-dlp sites (or a local path).

    142 GitHub stars~4k tokensUpdated 4 mo ago
    Media & CreativeAuto-check: warnings
  • Youtube Clipper

    op7418/Youtube-clipper-skill

    YouTube 视频智能剪辑工具。下载视频和字幕,AI 分析生成精细章节(几分钟级别), 用户选择片段后自动剪辑、翻译字幕为中英双语、烧录字幕到视频,并生成总结文案。

    2.2k GitHub stars~1.6k tokensUpdated 8 mo ago
    Media & CreativeAuto-check: notes
  • Video Transcribe

    wendy7756/AI-Video-Transcriber

    Transcribe and summarize a video or podcast from a URL (YouTube, TikTok, Bilibili, Apple Podcasts, SoundCloud, 30+ platforms) or from a local media/.txt file.

    3.3k GitHub stars~937 tokensUpdated 26 days ago
    Media & CreativeAuto-check: notes
  • Explainroo

    vincentsch/explainroo

    Make an explainer video (MP4) with a voice-over using explainroo.

    515 GitHub stars~439 tokensUpdated 7 days ago
    Media & CreativeAuto-check passed
  • Youtube Transcribe

    kennyzir/7deer_skills

    YouTube video transcription and memory workflow. An agent skill from kennyzir/7deer_skills.

    322 GitHub stars~1.7k tokensUpdated 12 days ago
    Media & CreativeAuto-check passed

More from HUANGCHIHHUNGLeo/claude-real-video

  • Claude Real Video

    HUANGCHIHHUNGLeo/claude-real-video

    Watch a video for the user. An agent skill from HUANGCHIHHUNGLeo/claude-real-video.

    2.2k GitHub stars~639 tokensUpdated 2 days ago
    Auto-check passed
  • Claude Real Video For Agents

    HUANGCHIHHUNGLeo/claude-real-video

    Install and use crv (claude-real-video) — a tool that lets any AI agent watch videos by extracting scene-aware keyframes, deduplicating them, and transcribing audio.

    2.2k GitHub stars~2k tokensUpdated 2 days ago
    Auto-check: notes

Questions about Claude Real Video

What does Claude Real Video do?

幫 Leo 看影片。當 Leo 丟影片連結(YouTube/IG/TikTok)或本機影片檔,要摘要、分析、拆解對標時使用——Claude 不能直接吃影片,先用這個工具抽關鍵幀+逐字稿+運鏡節奏+聲音/語氣/手勢時間軸再讀。. Claude Real Video is an agent skill from HUANGCHIHHUNGLeo/claude-real-video.

When should I use Claude Real Video?

Claude Real Video fits situations like: media & Creative work in your project.

How do I install Claude Real Video in Claude Code?

Run `npx skills add HUANGCHIHHUNGLeo/claude-real-video --skill claude-real-video -a claude-code`. Or copy the skill folder (personal-skill-backup in HUANGCHIHHUNGLeo/claude-real-video) into .claude/skills/claude-real-video in your project. Claude Code loads it when a task matches its description.

How do I install Claude Real Video in Codex?

Run `npx skills add HUANGCHIHHUNGLeo/claude-real-video --skill claude-real-video -a codex`. Or copy the skill folder (personal-skill-backup in HUANGCHIHHUNGLeo/claude-real-video) into .agents/skills/claude-real-video in your project. Codex loads it when a task matches its description.

Can I use Claude Real Video in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add HUANGCHIHHUNGLeo/claude-real-video --skill claude-real-video -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/claude-real-video, .gemini/skills/claude-real-video, .github/skills/claude-real-video and .opencode/skills/claude-real-video in your project.

What does Claude Real Video need to run?

Going by SKILL.md and its folder, Claude Real Video needs the command-line tools its instructions call (python3). Our summary lists: Python 3.

Does Claude Real Video access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Claude Real Video safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Claude Real Video use?

Claude Real Video is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Claude Real Video use?

About 559 tokens (SKILL.md is roughly 2.2k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Claude Real Video?

Skills that share tags, products or a category with Claude Real Video: Ffmpeg Skill (kajisho5/ffmpeg-skill, 1.9k stars), Watch (mathiaschu/watch, 142 stars), Youtube Clipper (op7418/Youtube-clipper-skill, 2.2k stars) and Video Transcribe (wendy7756/AI-Video-Transcriber, 3.3k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Claude Real Video?

HUANGCHIHHUNGLeo (a GitHub user) maintains it in HUANGCHIHHUNGLeo/claude-real-video, which has 2,209 GitHub stars. The repository holds 3 skills in this directory. The repository was last updated on October 8, 2026.

Source: HUANGCHIHHUNGLeo/claude-real-video on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.