ClawHub
Back to skill

Security audit

Venice Ai

Security checks across malware telemetry and agentic risk

Overview

This is a transparent Venice AI API wrapper, but users should understand that chosen prompts, files, media, and URLs may be sent to Venice or related search providers.

Install only if you are comfortable giving the skill a Venice API key and sending selected prompts, files, images, audio, video, and URLs to Venice AI for processing. Avoid internal or sensitive URLs, verify model-specific privacy terms before using confidential or regulated data, and use a dedicated API key with spending controls where possible.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (13)

Lp3

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding
The skill clearly instructs use of environment variables, local file inputs/outputs, and networked API calls, but the manifest does not declare permissions or otherwise surface those capabilities explicitly. This creates a transparency and governance gap: users or agent frameworks may invoke a highly capable external-service skill without clear consent boundaries for secret access, file handling, and outbound transmission.

Context-Inappropriate Capability

Low
Confidence
89% confidence
Finding
The helper reads credentials not only from the declared environment variable but also from a separate local application config under the user's home directory. That expands host-state access beyond the minimum needed for this utility and creates an unnecessary path for secret discovery if the skill is invoked in a broader automation context.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The top-level description markets the skill as a universal AI platform for nearly any task, with no scope limits, exclusions, or safe usage boundaries. Overly broad invocation language increases the chance an agent will select this skill for sensitive or unrelated workflows and then transmit data externally or perform unexpected file/network operations.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The manifest notes repeat a trigger-style 'one skill for everything' framing that encourages indiscriminate routing of user tasks into this external AI service. In agentic environments, such broad descriptions can cause over-selection and accidental exposure of data or capabilities beyond what a user reasonably expects.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The documentation actively encourages processing sensitive data with third-party models based on privacy claims, but it does not provide a prominent, balanced warning about residual risks, model/provider differences, account compromise, metadata leakage, or legal/compliance constraints. This may cause users or agents to send confidential material off-host under an overstated assumption of safety.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The X/Twitter search feature is presented as a normal capability, while the fact that it routes through xAI infrastructure appears only in a later note. Because prompts and search context may be sent to a different provider than the main Venice service, the lack of an upfront warning materially increases the risk of unintended third-party disclosure.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
The API explicitly supports scraping URLs found in user messages to augment model context, which can transmit user-supplied links and potentially linked content to external systems without an explicit privacy warning in the reference. In a skill advertising private inference, this is especially risky because users may assume URLs are not fetched or analyzed externally unless clearly told otherwise.

Missing User Warnings

Low
Confidence
72% confidence
Finding
The documentation requires a bearer API key but provides no guidance on secure storage, rotation, or avoiding accidental exposure in client-side code, logs, or repositories. This omission can lead integrators to mishandle credentials, resulting in unauthorized API use and billing abuse.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
When `--background-remove` is used with a local path, the script reads the local file and uploads its contents to the Venice API without an explicit warning at the decision point. In an agent-skill context, this can cause unintentional exfiltration of sensitive local images if a user or upstream agent supplies a path assuming processing is local.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The helper accepts a local file path, converts it to a data URL, and later sends it to a remote Venice API endpoint without any explicit user-facing warning at the point of use that local media contents will be uploaded off-host. In a CLI/agent skill context, this can cause unintended disclosure of sensitive local images, videos, or audio if the caller passes a local path assuming only local processing.

Missing User Warnings

Medium
Confidence
84% confidence
Finding
The transcribe command reads local audio files or fetches remote audio URLs, then uploads the content to Venice's remote API without an explicit user-facing notice at the point of use. In a skill context, this can cause users to unknowingly transmit sensitive voice content, recordings, or embedded metadata off-host to a third party.

Missing User Warnings

Medium
Confidence
87% confidence
Finding
The analyze command converts local images to base64 data URLs or accepts remote URLs, then sends that content to the Venice API for inference without an explicit warning that image data leaves the local environment. This is risky because users may provide screenshots, documents, IDs, or other sensitive images assuming local analysis.

External Transmission

Medium
Category
Data Exfiltration
Content
audio_url: str | None,
) -> dict:
    """Queue a video generation request."""
    url = "https://api.venice.ai/api/v1/video/queue"

    payload: dict = {
        "model": model,
Confidence
90% confidence
Finding
https://api.venice.ai/

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.