ClawHub

Vnsh Skill

Security checks across malware telemetry and agentic risk

Overview

The skill performs encrypted file sharing as described, but it gives agents too much automatic authority to fetch links, upload large outputs, and run an unpinned remote installer.

Install only if you are comfortable with vnsh.dev being used for external file sharing. Configure the agent to ask before opening vnsh links or uploading outputs, review and redact logs, diffs, source code, credentials, and context before sharing, and inspect or replace the curl-pipe installer with a safer verified installation path.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (6)

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The skill instructs the agent to automatically upload long outputs to an external service without asking the user first. Even if the content is encrypted, this still transmits potentially sensitive model outputs, user data, file contents, or internal context off-platform without consent, which is a clear data exfiltration risk.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill directs the agent to automatically decrypt and process any vnsh URL found in a message without user confirmation. This causes the agent to fetch and handle externally supplied content automatically, creating unnecessary exposure to untrusted inputs and potentially pulling sensitive or malicious material into the session.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The script uploads encrypted file contents to vnsh.dev, but it provides no explicit user-facing warning or confirmation at the point of execution that local file contents are being transmitted off-host. In an agent/skill context, this can lead users or upstream automations to disclose sensitive local data to a third-party service without sufficient awareness, even though the transfer is part of the stated functionality.

Ssd 3

High
Confidence
99% confidence
Finding
Automatic sharing logic can disclose user-provided data, generated analysis, logs, diffs, or confidential context to a third-party service without an approval step. The skill's framing as 'secure' and 'zero-knowledge' does not eliminate the disclosure event; it only changes the storage model, while metadata and possession of the link can still expose content.

Ssd 3

Medium
Confidence
93% confidence
Finding
Auto-reading any vnsh link makes the agent decrypt and process remote content solely based on pattern matching in chat. In context, this is more dangerous because the skill further instructs follow-on handling such as reading files or invoking image analysis, expanding the trust boundary from a pasted URL to arbitrary fetched content.

External Transmission

Medium
Category
Data Exfiltration
Content
# Encrypt and upload
RESPONSE=$(openssl enc -aes-256-cbc -K "$KEY" -iv "$IV" -in "$FILE" | \
  curl -s -X POST \
    --data-binary @- \
    -H "Content-Type: application/octet-stream" \
    "https://vnsh.dev/api/drop?ttl=$TTL")
Confidence
98% confidence
Finding
curl -s -X POST \ --data-binary

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal