NVIDIA Kimi Vision

ReviewAudited by ClawScan on May 1, 2026.

Overview

This skill appears to do what it says—send a chosen image and prompt to NVIDIA's vision API—but users should remember that images and an API key are involved.

This skill is coherent and purpose-aligned. Before installing or using it, make sure you are comfortable sending selected images and prompts to NVIDIA, protect the API key, and verify that the file path you provide is the image you intend to analyze.

Findings (3)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

Low

#ASI02: Tool Misuse and Exploitation

What this means

If the skill is pointed at the wrong local file, that file's contents could be uploaded as the image payload.

Why it was flagged

The script reads whatever local path is supplied and packages it for the API. That is necessary for image analysis, but there is no hard validation that the file is actually a supported image.

Skill content

image_path = sys.argv[1] ... with open(image_path, "rb") as f:
        b64 = base64.b64encode(f.read()).decode("utf-8")

Recommendation

Use the skill only with intended image files and confirm paths before analyzing private or sensitive files.

Low

#ASI03: Identity and Privilege Abuse

What this means

Anyone who obtains the API key may be able to use the associated NVIDIA API access or quota.

Why it was flagged

The skill uses a local NVIDIA API key file or provided key to authenticate requests. This is expected for the NVIDIA API, but it is still account-authorizing material.

Skill content

path = os.path.expanduser("~/.config/nvidia-kimi-api-key") ... headers={"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}

Recommendation

Store the key securely, avoid sharing it, and prefer a dedicated low-privilege API key if available.

Medium

#ASI07: Insecure Inter-Agent Communication

What this means

Private images, screenshots, or text in images will be transmitted to NVIDIA for processing.

Why it was flagged

The user's prompt and base64-encoded image are sent to NVIDIA's external API endpoint. This is the disclosed core function of the skill.

Skill content

INVOKE_URL = "https://integrate.api.nvidia.com/v1/chat/completions" ... "text": prompt ... "url": f"data:{mime};base64,{b64}" ... requests.post(INVOKE_URL,

Recommendation

Do not use the skill on sensitive images unless you are comfortable sending them to NVIDIA's API under NVIDIA's terms and retention policies.