← Back to Skills Marketplace
224
Downloads
3
Stars
0
Active Installs
6
Versions
Install in OpenClaw
/install mi-visionforge-svor
Description
Video Object Remove with SVOR(stable video object removal).
Usage Guidance
This skill appears to do what it claims, but it uploads your full video files to https://mipixgen-pre.ai.mioffice.cn for processing. Before installing/use: (1) only test with non-sensitive videos; (2) verify the origin and legitimacy of the SVOR_API_KEY and the service URL (the repo and domain are cited in SKILL.md but the skill package lists no homepage); (3) run in an isolated environment if you are unsure; (4) don't store the API key in shared or public places; (5) ensure python3, ffmpeg and the requests package are available; (6) review the remainder of the script (there was a truncated listing in the manifest you provided) to confirm no additional endpoints or behaviors are present. If you need higher assurance, ask the publisher for provenance or prefer a published release from a known repository.
Capability Analysis
Type: OpenClaw Skill
Name: mi-visionforge-svor
Version: 1.0.4
The MiVisionForge-SVOR skill bundle provides a legitimate tool for video object removal using the SVOR API. The core logic in `scripts/video_remove.py` uses `ffmpeg` for local video processing and the `requests` library to communicate with a Xiaomi-hosted service (mipixgen-pre.ai.mioffice.cn). While the skill involves uploading video files to a remote endpoint, this behavior is clearly documented in `SKILL.md` along with a privacy warning. The script handles the `SVOR_API_KEY` securely via environment variables and lacks any indicators of malicious intent, such as data exfiltration of sensitive local files or unauthorized command execution.
Capability Assessment
Purpose & Capability
Name/description, required binaries (python3, ffmpeg), and the single required env var (SVOR_API_KEY) align with a cloud-backed video object removal tool. The code posts videos and masks to the stated BASE_URL and uses ffmpeg for local processing — all expected for this purpose.
Instruction Scope
SKILL.md and the script instruct reading the input video, compressing/processing it, uploading it to the remote SVOR service, downloading masks/results, and merging masks locally. The script reads only the video and local temp files necessary for processing; no unexpected file-system or credential access is present in reviewed sections.
Install Mechanism
No install spec (instruction-only with an included script). This is lower risk; the script depends on existing python3, ffmpeg, and the requests Python package (not installed automatically). No remote arbitrary installers or extracted archives were used.
Credentials
Only SVOR_API_KEY is required as an environment credential, which is proportionate for authenticating to the remote SVOR service. The script does not request unrelated secrets or config paths.
Persistence & Privilege
Skill is not always-enabled and requests no system-wide persistence or modification of other skills. It runs as an on-demand CLI script and does not attempt to alter other agent settings.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install mi-visionforge-svor - After installation, invoke the skill by name or use
/mi-visionforge-svor - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.5
- 更新了服务运营商名称,由“小米Tools团队”改为“小米大模型应用团队”
- 隐私与安全声明中的相关团队名称同步更新
- 其他内容未变,功能与参数保持一致
v1.0.4
- Updated skill description to a concise English summary.
- Removed detailed usage instructions, background, and step-by-step guides from the skill manifest.
- Kept main functionality, dependencies, parameters, and privacy notes unchanged.
- No file or functional changes detected; documentation only.
v1.0.3
- 限制单次可消除视频长度由 5 秒调整为 3 秒
- 其余功能与参数说明保持不变
v1.0.2
- SKILL.md 描述明显标明只支持最长5秒视频内容消除。
v1.0.1
- 修复api_key传入代码逻辑
v1.0.0
MiVisionForge-SVOR v1.0.0
- Initial release: Implements stable video object removal (SVOR) skill via command line, powered by Xiaomi Tools team cloud service.
- Supports multi-class object segmentation and removal with VLM-based smart recommendations and flexible annotation (text/point/box).
- Requires transmitting videos to the Xiaomi-operated cloud for processing; results returned via expiring URL.
- Offers both stepwise (segmentation + removal) and one-step command-line workflows for various scenarios.
- Includes detailed usage instructions, parameter explanations, error codes, and privacy guidelines.
Metadata
Frequently Asked Questions
What is MiVisionForgeSVOR?
Video Object Remove with SVOR(stable video object removal). It is an AI Agent Skill for Claude Code / OpenClaw, with 224 downloads so far.
How do I install MiVisionForgeSVOR?
Run "/install mi-visionforge-svor" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is MiVisionForgeSVOR free?
Yes, MiVisionForgeSVOR is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does MiVisionForgeSVOR support?
MiVisionForgeSVOR is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created MiVisionForgeSVOR?
It is built and maintained by Fei Wang (@wangfei1204); the current version is v1.0.4.
More Skills