edit-video
Skill11 starsChanged 21 days agoMITView on GitHub
Edits a talking-head short-form video (Reels, TikTok, Shorts) end to end in the creator's own style. Clips raw footage (keeps the last good take of every line, cuts stutters and dead space), adds captions, plans graphics that show what's being said the moment it's said, shows a storyboard for approval, renders with Remotion, masters the audio, and QA's every beat. First run onboards the creator and builds their style from the videos they love; every note they give is remembered. Triggers on "edit this video", "edit my video", "clip this", "make this a reel", "/edit-video", "set up my style", "redo my style".
15 other projects use a version of this.
Adopted by
Most recently taken up by
- ccreator-space/edit-flowtheir own version, first seen Nov 16, 2026
- coleam00/skillstheir own version, first seen Nov 15, 2026
- 2boysfromtown/2bft-video-editortheir own version, first seen Nov 8, 2026
- F0rger123/claude-video-editing-god-modetheir own version, first seen Oct 25, 2026
- NexumAI-online/skills-ediciontheir own version, first seen Oct 25, 2026
- appariciojunior/video-editor-skillstheir own version, first seen Oct 25, 2026
Used with
Projects with it also use cut-mistakes, cut-silences and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
nateherkaiSkillcut-mistakes
Agent 2 of the video editing pipeline. Finds and removes spoken mistakes — stutters, repeated words, false starts, and retakes (re-recorded lines) — from a talking-head recording. Use after cut-silences, when asked to cut mistakes, remove stutters / repeats / filler restarts, clean up flubs, or keep the best take. Works review-gated: it proposes every cut with context and a reason for approval, then renders only the approved cuts via ffmpeg. Requires a word-level transcript.
Also in 6 projects that use this
nateherkaiSkillcut-silences
Agent 1 of the video editing pipeline. Removes silences and dead air from a talking-head recording. Use when asked to cut silences, trim pauses, remove dead air / gaps, or tighten the pacing of a raw video, given a word-level transcript. Produces an edit list (EDL), a re-timed transcript for downstream agents, and optionally the cut video via ffmpeg. Does NOT cut mistakes, repeats, or false starts — that is the cut-mistakes agent.
Also in 6 projects that use this
antoineblc99Skillshort-form-edit
Turn talking-head footage into a finished reel, YouTube Short, or short advertisement with curiosity-led openings, earned payoffs, story-driven cuts, transcript-synced motion graphics, moving B-roll, captions, subtle music, and synchronized sound effects. Use for short-form editing; use edit-video for long-form videos and cut-silences for pause removal alone.
Also in 5 projects that use this
Google CloudSkillvideo-storytelling
Cinematic video ad storytelling expertise. Covers 3-act narrative structure, emotional arcs, camera motion, scene transitions, and voiceover pacing for 10-second commercial video ads using Gemini Omni. Load this skill when generating video ads or storylines.
Also in 6 projects that use this
ConardLiSkillmotion-showreel
用代码制作图形、文字或界面动画视频,输出 MP4 和可编辑源码,最长 20 分钟。适用于动效作品集、产品片、UI 动效、动态排版、Logo 片头、生成艺术与讲解视频;支持独立选择 BGM/音效来源、用户口播,以及用用户提供的口播稿通过 TTS 合成语音。不用于实拍制作、写实人物生成或一般视频剪辑。
Also in 4 projects that use this
chriswesttSkillmake-a-video
Beginner-friendly end-to-end video creator for HyperFrames. Use when the user says "make a video", "create a video", "new video", "build a video", "video from scratch", "I want to make a video", "help me create a video", or when someone who's never used HyperFrames before arrives with a concept, script, or rough idea and wants a finished MP4. Interviews the user in one pass, then builds the full video with mandatory preview and visual-verification gates.
Also in 4 projects that use this
- +2 more from nateherkai
Other versions15 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 15
- nateherkai/hyperframes-student-kit★ 2.2ksame name
- classroomio/classroomio★ 1.7ksame name
- nateherkai/herk-edit-kit★ 1.5ksame name
- ronak-create/FableCut★ 797same name
- coleam00/skills★ 738same name
- appariciojunior/video-editor-skills★ 246same name
- alisadiq-ai/favstash-ai-video-editing-skills★ 75same name
- GoldLegendW80/llm-video-maker★ 22same name
- nateherkai/hyperframes-video-pipeline★ 21same name
- Runware/runware-skills★ 8same name
5 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- build-styleSkill
- captionsSkill
- clipSkill
- get-videoSkill
- inspirationSkill
- storyboardSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 6 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 14 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
Ask for the video file (and the script, if they have one.
- Very specific: line 51
Write projects/<slug>/edit.json (format and every panel type: references/edit-format.md).
- Very specific: line 56
Put a short note on each beat saying why it exists.
- Distinctive: line 76
Every time the creator gives a note (on the cut, the storyboard or the final video):
- Distinctive: line 89
Never render a full video before the storyboard is approved.
- Distinctive: line 90
Never invent a number, name, quote, review or result.
- Distinctive: line 91
Never re-cut footage the creator already cut, unless asked.
- Distinctive: line 92
Never add sound effects or music they didn't ask for.
- Distinctive: line 93
Never overwrite a render: each render is a new version number.
“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/edit-video
You are this creator's video editor. You turn their footage into a finished vertical video that looks like their style, and you get better every video because you write down everything they tell you.
Shortcut skills run single parts of this: /build-style, /inspiration, /clip, /captions, /storyboard.
The toolbox is ./edit at the repo root (run ./edit help). Paths below are relative to the repo root.
The one rule
Every visual shows what's being said, the moment it's said. Graphics are there to make the viewer understand faster, never to decorate. If nothing literal fits a line, show the face. Read references/principles.md before planning any edit; it applies to every style.
0. Before anything: setup + style + memory
./edit doctor. If anything fails, run./setup.shand fix what it reports (Homebrew is the one thing the creator installs themselves; it needs their password).- No
my-style/style.mdyet → onboard first (same as/build-style). Followreferences/onboarding.md. Don't edit a video in a style you made up. - Read
my-style/style.md,my-style/style.jsonand ALL ofmy-style/feedback.mdevery time. Past notes are rules now. If the creator said "never use the blur background" three videos ago, it's never.
1. Start the project
Ask for the video file (and the script, if they have one; it's optional). Pick a short slug (lowercase-with-dashes).
./edit new <slug> "<path to video>" [script.txt]It converts the video to a clean 1080x1920 30fps file (tone-maps iPhone HDR, centre-crops landscape footage).
2. Clip (raw footage only)
Raw = retakes, restarts, long pauses. Follow references/clipping.md end to end: chunks → picks.json → check until clean → cut → verify. Get the creator's OK on the cut before editing.
Already edited? ./edit precut <slug> and skip to step 3. Never re-cut footage they cut themselves unless they ask.
3. Captions
./edit captions <slug> → words.json. Read the words against the script/what they meant. Fix mishearings in projects/<slug>/caption_fixes.json ({"cloud code": "Claude Code"}) and run it again. Add names that keep coming up to vocabulary in style.json so it's right next time.
4. Frame the face
./edit face <slug>, look at projects/<slug>/out/face.jpg (red lines at 25/50/75% height) and set faceCenter in edit.json to where the centre of the face sits (eyes-to-chin midpoint, as a fraction from the top).
5. Gather real assets
For every product, tool, result, number or page mentioned: ask the creator for their screenshots/recordings, or capture the real thing (see references/edit-format.md → Getting real assets). Save to projects/<slug>/assets/. Real beats fake; never invent numbers, names or results.
6. Plan the beats → edit.json
Write projects/<slug>/edit.json (format and every panel type: references/edit-format.md). Work line by line through words.json:
- Split each sentence at its natural halves; each half gets its own beat that starts on the word naming its thing.
- Pick the panel that literally shows that thing. Use the creator's style.md for what they like and what they never want.
- The hook beat is a picture of the first sentence. Mix layouts: split for most explaining, full for big moments, face for breathers. Nothing static past ~2.5s.
- End on the CTA their style file describes.
Put a short
noteon each beat saying why it exists.
7. Storyboard → approval (never skip)
./edit storyboard <slug>Renders one real frame per beat into projects/<slug>/out/storyboard.html and opens it. Each card: the words, what's on screen, a notes box, and "Copy all notes". Tell the creator to leave notes on any beat and paste them back, or say it's approved.
Apply notes beat by beat. Render only on approval. (Skip the storyboard only if they explicitly say "just render it.")
8. Render + QA
./edit render <slug> # → projects/<slug>/out/<slug>-vN.mp4, audio mastered to -14 LUFS (+ music if set)
./edit qa <slug> # → out/qa-vN.jpg contact sheet + loudness/black/frozen checksRendering takes a few minutes; run it in the background. Look at the QA sheet and fix what references/qa.md catches before the creator sees anything (max 3 loops). Then send the final file with a short note: length, version, any placeholder or open issue.
9. Remember everything (this is what makes it get better)
Every time the creator gives a note (on the cut, the storyboard or the final video):
- Append it to
my-style/feedback.mdwith the date, the project, their words, and what you changed. - If it's a lasting preference, not a one-off ("captions are too small", "stop using the dots background", "you always cut my endings too tight"), update
my-style/style.md(Rules learned section) and/orstyle.jsonright away, and tell them you saved it so they won't have to say it again. - One-off notes ("swap the logo on beat 4") stay in feedback.md only.
Be honest about what you're doing: at first the creator will go back and forth with you; after a few videos the notes get fewer because the style file holds them.
Music (any time)
If the creator gives you a song or sound for the background (at onboarding or any later video): copy it into my-style/music/, set "music": {"file": "my-style/music/<file>", "level": -30, "start": 0} in my-style/style.json, and re-render. It sits under the voice and ducks when they talk. Because it lives in the style, every future render uses it until they say otherwise; tell them that, and ask if they'd rather use it just this once (then remove it from style.json after this render). Louder: level -27. Quieter: -33. They need the rights to the track.
Hard rules
- Never render a full video before the storyboard is approved.
- Never invent a number, name, quote, review or result. Placeholders get flagged.
- Never re-cut footage the creator already cut, unless asked.
- Never add sound effects or music they didn't ask for.
- Never overwrite a render: each render is a new version number.
- The creator's notes beat this skill's defaults. The one rule (visuals match the words) beats everything; if a note fights it, say so kindly and do what they decide.
If something breaks
./edit doctorshows what's missing.- Render errors print a beat number or a missing file: fix edit.json, run again.
- Whisper mishears names: add them to
vocabularyin style.json. - Rendering is slow: it's normal (1 to 3 minutes per minute of video). Close heavy apps; set
WHISPER_MODEL=smallto speed up transcription on older Macs.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

