ClawHub

agent-init

Security checks across malware telemetry and agentic risk

Overview

This skill is a user-directed OpenClaw workspace initializer with one disclosed setup risk: an optional command that installs uv by running a remote script.

Install this only if you want an agent to help create or update OpenClaw workspace instruction files. Review every generated file before approving writes, keep secrets out of workspace markdown, verify the correct host/container path, and prefer a trusted package manager or reviewed installer over the optional curl-piped uv install command.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Context-Inappropriate Capability

Medium
Confidence
97% confidence
Finding
The skill directs use of `curl ... | sh` to fetch and immediately execute a remote installer script. This bypasses normal integrity review and grants arbitrary code execution from a network resource, which is significantly broader and riskier than merely configuring workspace markdown files.

Context-Inappropriate Capability

Medium
Confidence
98% confidence
Finding
The script executes a remote installer directly with `curl ... | sh`, which allows arbitrary code from the network endpoint or any upstream compromise to run immediately in the user's shell. In an agent initialization skill, this is especially risky because users may run it during setup with elevated trust and possibly broad local permissions.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal