ClawHub
Back to skill

Security audit

AWI

Security checks across malware telemetry and agentic risk

Overview

AWI is a disclosed web-reading and search skill, but installing it means trusting a downloaded GitHub release binary.

Install only if you trust the jzOcb/awi GitHub release binary, because the installer downloads native code without checksum or signature verification. Avoid using it with secrets, private internal URLs, confidential searches, or regulated data unless you are comfortable with those requests leaving your environment.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding
The skill is presented primarily as a web-reading/search utility, but it also installs software by running `scripts/install.sh`, downloads a binary from GitHub releases, and encourages executing that binary. This hidden operational behavior expands trust requirements substantially and can expose users to supply-chain compromise or unintended code execution, especially because the install path is framed as simple setup rather than a security-sensitive action.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill advertises web reading and search, which necessarily sends user-provided URLs, queries, and potentially related metadata over the network, but it gives no warning about external requests or data exposure. In an agent context, users may pass sensitive links, search terms, or internal targets, making silent transmission more dangerous than in a normal browser-like tool.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The installation instructions tell the user to run a shell script that installs software, but they do not clearly warn that this action downloads and runs code on the user's machine. In a skill ecosystem, such instructions normalize execution of installer scripts and raise the risk of supply-chain attacks, local compromise, or persistence if the source or distribution channel is tampered with.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.