Search skills/

OpenMontage agent skills

An agentic video production system with 140 skills and 100+ tools that lets a coding agent script, source assets, edit and render finished videos.

calesthio/OpenMontage 140 skills Video & media

with caveats

A deep, actively developed video stack, but you need to know before starting that it requires Python and Node setup, that most generative output runs through paid third-party APIs, and that the code is AGPL-3.0.

broad access: pre-approves shell commands · installs packages → security details

Depth
4/5
Last update
4 days ago
Installs
—
Stars
74K
Forks
9.7K
License
AGPL-3.0, check terms

What OpenMontage does

OpenMontage lets a coding agent act as a video production crew. You describe a video in plain language, and the agent handles research, scripting, asset generation, editing and final composition. The README lists 12 production pipelines. One of them, Documentary Montage, cuts real motion footage from a CLIP-indexed corpus of free stock and open archives (Pexels, Archive.org, NASA, Wikimedia, Unsplash), so you can produce edited video without paid models.

The 140 skills in .agents/skills/ average about 8.5 KB each. Each covers one tool or provider: FFmpeg, ComfyUI, Remotion and HyperFrames composition, Blender 3D worlds, ACE-Step music on RunPod, ElevenLabs voice agents, Azure speech, BFL image generation and AI video models such as Kling, Veo and Seedance. Many skills point the agent at Python tools in the repo, for example tools/music_gen.py. Some include reference folders or scripts. The 3d-asset-generation skill has a paid-call rule: before the first call to a provider, the agent states the provider, model, estimated unit cost and number of outputs, then generates one sample before running a batch. The repo also has CI, contract tests that run without API keys, and an AGENT_GUIDE.md that defines the agent contract.

Good for

  • Developers who want an agent to produce short narrated or music-driven videos end to end
  • Documentary-style montages built from free stock and archival footage
  • Projects already using Remotion, HyperFrames, Blender or FFmpeg

Not for

  • Anyone wanting a lightweight skill to drop into an existing project
  • Teams that cannot accept AGPL-3.0 terms
  • Users who want generative video without managing paid provider keys

What's inside depth 4/5

140 skills averaging 8.5 KB, plus reference folders, helper scripts, 100+ Python tools, CI and 8 eval files across 2,209 files, though most individual skill folders have no tests of their own.

Skills
140
Reference docs
800
Scripts
246
Tests / evals
18
Templates
407
Avg SKILL.md
8.3 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Python 3 with a virtualenv and requirements.txt
  • Node.js and npm for the remotion-composer project
  • piper-tts for local narration
  • FFmpeg; Blender for 3D pipelines
  • Optional paid API keys: Atlas Cloud, Kling, Suno, ElevenLabs, OpenAI, xAI, Google, Volcengine Ark, HeyGen, Runway, BFL, fal.ai
  • RunPod account and endpoint for ACE-Step music generation
  • Free Pexels and Pixabay API keys for stock footage

Security review

Security: Broad access

This skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.

  • Pre-approves broad shell commandsFrontmatter allowed-tools: Bash, Read, Write pre-approves unrestricted shell for this skill..agents/skills/gemini-omni/SKILL.md:5

This is a large video-production skill collection that calls many paid media APIs (HeyGen, fal.ai, ElevenLabs, BFL, Atlas Cloud, Azure, RunPod and others), using each vendor's own API key and uploading the inputs you provide for generation. The gemini-omni skill pre-approves unrestricted Bash. Some workflows auto-install unpinned npm packages or rely on external CLIs, one of which (HyperFrames) sends telemetry you can disable. Most of the repo's files, including its Python tools, were not available for review.

Worth knowing

  • Installs packages or toolsIf the beautiful-mermaid package is missing, the script installs it with npm install or bun add..agents/skills/beautiful-mermaid/scripts/render.ts:157
  • Runs remote code at an unpinned versionensurePackage installs and imports beautiful-mermaid with no version pin..agents/skills/beautiful-mermaid/scripts/render.ts:138
  • Runs remote code at an unpinned versionThe HyperFrames workflow runs npx hyperframes ... with no version pin..agents/skills/hyperframes-cli/SKILL.md:3
  • Sends usage data by default (can be turned off)The HyperFrames CLI sends anonymous usage telemetry, which can be turned off with npx hyperframes telemetry disable..agents/skills/hyperframes-cli/references/upgrade-info-misc.md:54
  • Installs packages or toolsInstructs a global install of @elevenlabs/cli with npm install -g..agents/skills/agents/SKILL.md:21
  • Relies on a CLI or package we could not reviewRelies on the ElevenLabs CLI (elevenlabs auth login, elevenlabs agents push), whose source is not in the repo..agents/skills/agents/SKILL.md:22
  • Relies on a CLI or package we could not reviewPNG capture relies on the agent-browser CLI..agents/skills/beautiful-mermaid/SKILL.md:98
  • Relies on a CLI or package we could not reviewMusic generation runs tools/music_gen.py on RunPod; that script was not among the files we have..agents/skills/acestep/SKILL.md:8
  • Needs an API key for its own serviceHEYGEN_API_KEY is sent only to api.heygen.com; other skills use FAL_KEY, KLING_API_KEY and GEMINI_API_KEY for their own gateways..agents/skills/ai-video-gen/SKILL.md:48
  • Needs an API key for its own serviceELEVENLABS_API_KEY is sent as xi-api-key to api.elevenlabs.io..agents/skills/agents/SKILL.md:87
  • Needs an API key for its own serviceBFL_API_KEY is sent as x-key to api.bfl.ai..agents/skills/bfl-api/references/code-examples/curl-examples.sh:33
  • Needs an API key for its own serviceAZURE_SPEECH_KEY is used for the user's own Azure Speech resource..agents/skills/azure-speech-to-text/SKILL.md:5
  • Needs an API key for its own serviceATLASCLOUD_API_KEY is used for the Atlas Cloud gateway..agents/skills/atlas-cloud/SKILL.md:13
  • Makes network requestsCalls declared media-generation APIs (HeyGen, fal.ai, Kling, Gemini, Atlas Cloud, BFL, ElevenLabs, Azure, RunPod) and downloads the generated outputs..agents/skills/ai-video-gen/SKILL.md:133
  • Makes network requestsLocal input files are uploaded to Atlas Cloud as part of each generation request..agents/skills/atlas-cloud/SKILL.md:30

What we checked

The repository (2209 files) on Nov 1, 2026. Not a full review: 1769 of 2209 files were not downloaded (large repository; we reviewed the skill folders); we read 5 of 10 code files; it relies on code outside the repository: 1769 readable repo files not provided, including most skill reference docs and the OpenMontage Python tools (tools/music_gen.py, tools/sfx.py, tools/tool_registry, selectors and provider tools); .agents/skills/hyperframes-animation/scripts/animation-map.mjs; .agents/skills/hyperframes-animation/scripts/package-loader.mjs.

NVIDIA SkillSpector static scan: 556 pattern hits (Data Exfiltration 363, MCP Rug Pull 162, Memory Poisoning 15, Privilege Escalation 7, Excessive Agency 3, Supply Chain 2). The Data Exfiltration hits are documented curl/fetch/requests examples that call each vendor's own API (HeyGen, ElevenLabs, BFL) with that vendor's key. The MCP Rug Pull hits come from mcp__heygen__* tool references, and the Credential Access hits are instructions to store BFL_API_KEY in .env. The 'Concealed Executable Artifact' hits on the beautiful-mermaid .ts files are ordinary readable TypeScript scripts.

A review, not a certification. Skills change after we check them — read the files before installing.

License

AGPL-3.0: commercial use allowed, but if you modify it and offer it over a network you must publish your source. Some vendored skills, such as the ElevenLabs agents skill, carry their own MIT notice.

Can it run unattended?

Not really. The pipelines are multi-step, but the skills tell the agent to state costs and generate one sample before any paid batch, so a typical run has human checkpoints rather than an unattended loop.

Alternatives to OpenMontage

HyperFrames

heygen-com/hyperframes

HeyGen's framework and skills for writing HTML and rendering it to deterministic MP4 video, with a CLI, a block registry and media tooling.

with caveats Security: Broad access depth4/5

Remotion Agent Skills

remotion-dev/skills

The Remotion team's twelve skills for building videos as React code: project setup, markup rules, maps, captions, studio, render and upgrades.

recommended Security: Check first depth3/5

video-shotcraft

Vincentwei1021/video-shotcraft

Remotion-based skill for cinematic product promo videos, with 124 shot recipe cards, a full template, beat sync, sound design and a motion workbench.

with caveats Security: Check first depth5/5

Vox Director

Alisa0808/vox-director

A skill that turns one topic, a talking-head video or a photo into a finished paper-collage explainer video, using the Atlas Cloud API and local ffmpeg.

with caveats Security: Check first depth4/5

FAQ

What is OpenMontage?

OpenMontage is an open-source video production system for AI coding assistants. It ships 140 agent skills and over 100 tools so the agent can research, script, generate or source assets, edit and render a finished video. The README lists 12 production pipelines.

How do I install OpenMontage?

In the cloned repo, run make setup. On Windows, the README gives the manual steps: create a Python venv, install requirements.txt, run npm install in remotion-composer, install piper-tts and copy .env.example to .env. If npm fails on Windows with ERR_INVALID_ARG_TYPE, the README suggests npx --yes npm install.

Do I need paid API keys to use OpenMontage?

No. The README says that with zero keys you can still make real videos, for example the Documentary Montage pipeline built from free stock and archival footage. Adding keys for providers like Kling, ElevenLabs, OpenAI, Google or Runway enables more tools. Pricing and free tiers are in docs/PROVIDERS.md.

How much does an OpenMontage video cost to make?

It depends on the providers you use. The README's examples include a 60-second animated short made with Kling clips via fal.ai and Google narration for $1.33, and a 50-second vertical film for about $4. Free-footage pipelines can cost nothing beyond setup.

Is OpenMontage safe to install?

We found no hidden network calls or obfuscated code in the scanned material. Scanner hits come from documented provider API calls and instructions for storing API keys in .env. The main risk is spending: agents can make paid generation calls with the keys you supply, so set provider budgets where you can.

Curated by , AI enthusiast. Reviews are written by AI reviewers under our rubric and checked against the repository files.

Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.