Motion Graphics Skills
Professional motion, animation, and video workflows for AI coding agents.
- Source repo
- iart-ai/motion-skills
- Stars
- ★ 515
- Last updated
- 3d ago
- License
- MIT
- Primary language
- HTML
- FA score
- 39/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- TikTok, Reels, or Shorts creators can add the short-form pack when they want a coding agent to produce vertical videos.
- Not a fit if
- Teams seeking an all-in-one hosted rendering service
- Users wanting every capability in one installation
- Teams needing workflows outside the listed motion domains
- Source review
- 39/100 · Major gaps 4 safety controls not found
What does this agent do, and when should you use it?
Motion Graphics Skills is a collection of 53 open-source skills organized into 16 separately installable packs for distinct audiences and motion disciplines. Each skill is a self-contained directory containing `SKILL.md` and `references/`, teaching a coding agent a specific workflow rather than supplying a standalone editor or model. Its scope includes short-form video, animated data charts, explainers, kinetic typography, web animation, WebGL, Manim, Remotion, and frame-by-frame JavaScript Canvas work. Skills that create visual artifacts use a deliver-and-verify loop and may include scripts for freezing frames, assembling contact sheets, and probing encoded MP4 files. Web-oriented skills produce standalone HTML, while video skills render through Remotion or Manim; installed skills are documented as auto-discoverable by Claude Code, Codex, Cursor, and more than 40 agents.
A user selects one of the 16 subject-specific packs and installs its repository with npx skills add or, for Claude Code, through the plugin marketplace. The host agent reads the pack's SKILL.md and references/ and follows the encoded professional workflow to build the requested motion artifact. For visual work, supporting scripts/ can freeze a rendered frame, tile a contact sheet, or probe an encoded MP4, while the workflow asks the agent to render, capture, and inspect its result. These checks address timing, easing, safe areas, typography, color, frame accuracy, and seamless loops. Depending on the skill, the resulting artifact is either standalone HTML or video rendered with Remotion or Manim; the repository supplies skills and verification tooling rather than a standalone hosted renderer.
- TikTok, Reels, or Shorts creators can add the short-form pack when they want a coding agent to produce vertical videos.
- Analysts and product managers can use the data-animation pack for frame-accurate animated charts and infographics.
- Educators can choose explainer or Manim skills for instructional clips and mathematical animation.
- Frontend developers can install web or JavaScript animation packs for GSAP, SVG, Lottie, Canvas, and other browser motion.
- Brand, advertising, and e-commerce teams can select the corresponding packs for product demonstrations and branded motion creative.
- Motion designers can use the motion-design and kinetic-typography packs for titles, logos, type animation, and brand-system motion.
How do you install or deploy this agent?
Packs are installed separately; no command for installing all 16 packs at once is documented. The provided examples are:
npx skills add iart-ai/tiktok-video-skills
npx skills add iart-ai/data-animation-skillsA pack can also be added as a Claude Code plugin:
/plugin marketplace add iart-ai/web-animation-skillsA terminal capable of running npx and network access are needed for these installation paths. The source does not state a minimum Node.js version, credential requirements, or a shared runtime configuration.
How do you use this agent?
After installing a pack suited to the task, supported coding agents such as Claude Code, Cursor, and Codex auto-discover its skills and activate them when a prompt matches the work. The selected skill guides artifact construction and, for visual outputs, applies the render-a-frame, capture, and inspect verification loop. Web skills emit standalone HTML, while video skills render with Remotion or Manim. The source does not provide a canonical first prompt, one shared generation command, or a common output path, so the exact invocation depends on the installed skill and host agent.
What are this agent's strengths and limitations?
- The 53 skills are divided into 16 audience-specific packs, allowing users to install only the relevant workflows.
- Visual skills include a deliver-and-verify loop plus tools for frozen frames, contact sheets, and encoded MP4 inspection.
- The collection spans Remotion, Manim, Three.js/WebGL, GSAP, SVG, Lottie, and JavaScript Canvas workflows.
- Skills are self-contained
SKILL.md, reference, and script directories, with no model fine-tuning or dedicated plugin runtime required. - The MIT license permits broad reuse.
- This is not a standalone editor or complete hosted pipeline; execution depends on a host coding agent such as Claude Code, Codex, or Cursor.
- The 16 packs live in separate repositories and are installed individually, with no documented install-all command.
- Video workflows still depend on the selected skill's Remotion or Manim path, while web and video outputs lack a unified runtime.
- Minimum Node.js versions, Remotion and Manim setup, system requirements, and shared failure-recovery procedures are not documented in the supplied material.
- Although compatibility with more than 40 agents is claimed, only Claude Code, Cursor, and Codex are named, without a platform-by-platform compatibility matrix.
How does this agent compare with similar options?
Compared with iart.ai's full pipeline, this repository provides open-source skills and verification workflows that users bring into an existing coding agent. iart.ai is presented as the AI motion agent that can run the whole process from a prompt, CSV, or brand kit.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Motion Graphics Skills This agent | 39 · Major gaps | Agent plugin / skillFree | ★ 515 | 3d ago | HTML | Codex · Claude Code |
| Video Shotcraft | 0 · Major gaps | Agent plugin / skillFree | ★ 9.6k | 3d ago | TypeScript | Codex · Claude Code |
| OpenMontage Video Studio | 48 · Major gaps | CLIFree + model costs | ★ 61k | 21d ago | Python | Codex · Claude Code |
| FireRed OpenStoryline | 57 · Major gaps | CLIFree + model costs | ★ 3.4k | 1mo ago | Python | Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The README encourages installing only the needed pack and states that no plugin runtime is involved, offering limited evidence of least-privilege design; it provides no permission manifest or enforced isolation. No user-confirmation policy is documented for rendering, dependency installation, or file replacement. Data flow is only partially visible through scattered references to CSVs, brand kits, rendering, and screenshot inspection, with no account of storage, transmission, or third-party processing. Sensitive-data handling, credential protection, and dependency-security policies are absent. Installation and media generation imply local or external effects, but network access, write locations, and effect boundaries are not fully disclosed. No rollback or uninstall procedure is supplied. The builder, project affiliation, linked pack repositories, and copyright holder are attributed clearly, but publisher identity is not independently verified and no fuller contributor or provenance record is provided.
The 16 listed packs total 53 skills, matching the README heading, and the installation and product descriptions are internally coherent. Commands are provided for npx and Claude Code, and tools such as Remotion and Manim are named, but versions, prerequisites, lock data, and availability evidence are absent. No failure modes, diagnostic messages, error examples, or recovery guidance are documented.
The 16 packs are mapped clearly to creators, developers, analysts, educators, advertisers, and other concrete scenarios, earning full credit for audience and scenario coverage. Boundaries are partly defined by describing skills as SKILL.md-based knowledge packages and distinguishing web, video, WebGL, and Manim workflows, but per-pack inputs, outputs, and non-goals are not specified. Automatic activation is asserted only as matching the user's prompt; trigger conditions, exclusions, and conflict handling are missing. Compatibility is claimed for Claude Code, Cursor, Codex, and more than 40 agents, but only two installation patterns are shown and operating-system, runtime, and agent-version details are absent.
The README has a clear progression through pack catalog, installation, skill definition, use cases, and licensing, but repository-level navigation does not cover maintenance, troubleshooting, or contribution material. Installation commands are concise and support selective installation, while prerequisites, verification, upgrades, and removal are omitted. Pack names consistently use purpose-oriented “-skills” naming, although no compatibility or deprecation policy establishes long-term stability. A few commands and workflow descriptions serve as examples, but there is no complete tutorial, expected-output walkthrough, or FAQ. Known limitations are not documented. The complete MIT text matches the README declaration, justifying full license credit. No version scheme, release history, or changelog appears. iart.ai is identified as builder and copyright holder, but maintainer contacts, issue policy, and update path are not supplied.
The source describes editable artifacts, standalone HTML, Remotion or Manim video, and delivery checks using rendered frames, screenshots, contact sheets, and MP4 probing, giving useful expectations for outputs; the supplied files do not include the skills, detailed sample outputs, or quality criteria needed for stronger support. Encoded professional workflows and domain-specific packs may add value, but this is primarily asserted marketing language and the evidence does not demonstrate improvement over general prompting. Selective installation, no fine-tuning, no plugin runtime, and MIT licensing reduce adoption cost, while rendering-toolchain, compute, and dependency-maintenance costs remain unquantified.
The headline count can be checked against the table, and pack links provide traceability targets; claims about professional practice, support for more than 40 agents, automatic discovery, and verification effectiveness are not tied to implementation, tests, or specifications in the supplied material. The LICENSE corroborates the README's MIT declaration and iart.ai attribution, but no code, tests, skill inventory file, or independent documentation corroborates the central capability claims. Mechanistic statements are mixed with promotional language without consistently distinguishing implemented facts, intended behavior, and inference.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- This review covers only the README and LICENSE; no SKILL.md files, scripts, dependency manifests, tests, or downstream pack contents were supplied, so actual behavior and safety cannot be confirmed.
- The installation commands may download and execute third-party packages; inspect each target repository, pin a revision, and verify dependencies and install scripts before use.
- Automatic discovery, support for more than 40 agents, and the effectiveness of the delivery-verification loop are not demonstrated by the supplied implementation evidence.
- Sensitive-data handling, network access, file writes, overwrite confirmation, uninstall, and rollback are undocumented; establish data boundaries before supplying CSVs, brand kits, or other private material.
- The object description says 50 skills and 14 packs, while the README says 53 skills and 16 packs. The README's own count is consistent, but the external metadata and reviewed revision appear version-skewed.