메뉴
Orientation skill for new agents and users. Explains what Montaj is and how it works. Use it when the user asks questions about montaj or is getting started
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
You MUST use this whenever the user asks for video editing work. Use it when video-related tasks are brought up. Editing, analyzing video, or transcribing videos
Find and download real images from the web — people, logos, brand/event stills, B-roll — to add as overlay image cards or project assets. Load when the prompt asks to source or insert images (e.g. 'add a photo of X', 'find images of the IPO', 'pull a shot of the factory').
Agent-authored workflow task: analyze transcripts across all clips, pick ONE best take per script section, discard all others. Load this when you hit montaj/select_takes in a workflow.
Write a custom JSX overlay component and add it to the project's overlay track.
Agent-authored workflow task: transcribe-and-sample one long-form horizontal source, pick N self-contained clip windows, and fan each out into its own vertical (9:16) clip project with the chosen framing. Load this when you hit montaj/find_clips in a workflow.
The shared vocabulary that domain skills use to name Montaj operations. Domain skills phrase every Montaj interaction as one of these verbs; an interface skill defines how each verb is actually performed. Load this to learn the canonical verbs before authoring or reading a domain skill.
| name | onboarding |
| description | Orientation skill for new agents and users. Explains what Montaj is and how it works. Use it when the user asks questions about montaj or is getting started |
A video editing toolkit that mounts on top of an agent harness. You bring the brain. Montaj gives agents the tools.
Montaj doesn't edit on its own. It provides discrete operations — trim, transcribe, remove fillers, caption, resize, composite overlays — and the agent decides which to call, in what order, with what parameters. The agent is the editor. Montaj is the toolbox.
Most programmatic video tools (Remotion, Revideo) have the agent write code that describes a composition. Montaj takes the opposite approach: the agent calls tools against existing footage. Agent can still write code on top of existing footage for overlays and animations, but montaj is desigend to work around and with existing footage. The agent reasons about which steps to apply and executes them.
GUI tools (Premiere, DaVinci, Descript, Runway) are driven by a human. Every decision is a click. Montaj is designed for the agent to drive — the human reviews the result and performs last mile edits, the bulk of the process is agent driven.
Where Montaj is the right choice:
Where Montaj is not the right choice:
A project is one video being made. Everything lives in a single file: project.json.
pending → agent edits → draft → human reviews → final → render → MP4
pending — created, clips and prompt attached, waiting for the agentdraft — agent is done; ready for human reviewfinal — reviewed and approved; ready to renderThe agent reads project.json, calls steps, writes it back as it works. The UI watches the file and rerenders the timeline on every write.
~/Montaj/ ← workspace — all projects live here
2025-04-01-product-demo/
project.json ← the project (pending → draft → final)
clip.mp4
clip_spec.json ← trim spec (editing decisions, no re-encode)
clip_transcript.json ← word-level timestamps
overlays/hook.jsx ← agent-authored overlay components
output.mp4 ← final render
~/.montaj/ ← user config + extensions
config.json ← workspaceDir, model, render settings
credentials.json ← API keys (never committed)
profiles/<name>/ ← account style profiles
steps/ ← custom steps
workflows/ ← custom workflows
montaj/ ← built-in (ships with Montaj)
steps/
workflows/
skills/
A style profile captures the visual and editorial identity of a social media account — pacing, cut frequency, color palette, caption style, tone. Once created, it gets injected into every project for that account so editing decisions stay consistent automatically.
Profiles live at ~/.montaj/profiles/<name>/style_profile.md and are loaded into your agent context when tagged selected in the UI or manually seeked on the prompt.
To create or update one: load skills/style-profile/SKILL.md.
montaj serve starts a local server and opens http://localhost:3000.
The Editor tab drives the core flow: upload clips + prompt → watch the agent work live → review and tweak → render. The agent writes project.json as it works; the UI reflects every change in real time via SSE.
Four tabs: Editor, Workflows (node graph), Overlays (live JSX preview), Profiles (style profiles).
See docs/UI.md for the full breakdown.
Start the server using montaj serve. If you get warnings or errors, bubble them up to the user with the easiest way to fix them (running montaj install ui, etc.).
Ask the user:
"Before we start — do you want to set up a style profile for your account first, or jump straight into editing a video?
Style profile: I'll analyze your existing content (or a URL you paste) and build an editing style that gets applied automatically to every project for that account — pacing, caption style, color palette, tone.
Edit a video: Tell me the location of your clips or drop them in the UI, and I'll run the full edit pipeline for whatever content style you pick. You can always add a style profile later."
skills/style-profile/SKILL.md and follow it.skills/SKILL.md and follow it.