ClawHub

错敏信息检测

Security checks across malware telemetry and agentic risk

Overview

The skill does what it claims: it sends user-provided text or fetched webpage content to UCAP for sensitive-information checking, with the main privacy and URL-fetching risks disclosed.

Install only if you are comfortable sending the text or webpage content you ask it to inspect to UCAP. Avoid using it on secrets, private documents, source code, or regulated data unless UCAP is approved for that data. Leave dynamic browser mode disabled unless you intentionally configure a tight domain allowlist and understand the added browser-rendering risk.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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)

Context-Inappropriate Capability

Medium
Confidence
85% confidence
Finding
The skill provides broad URL fetching and optional browser-driven retrieval even though no manifest purpose is provided to justify those capabilities. In a security-sensitive skill that sends page contents to an external scanning API, this enlarged capability surface increases risk of misuse, unintended data collection, and network pivoting if protections are weakened or misconfigured.

Context-Inappropriate Capability

High
Confidence
90% confidence
Finding
Executing an external browser automation CLI is a high-risk capability because it can render active web content and initiate additional network activity outside normal HTTP fetching. Even with some SSRF controls, browser behavior is complex and can be abused through redirects, embedded resources, or future configuration drift, making this unjustified capability notably dangerous in an agent skill.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill transmits provided text or fetched webpage contents to a third-party API for sensitive-information analysis without any user-facing warning, consent, or minimization. Because the very purpose of the skill is to process potentially sensitive material, silent external transmission materially raises privacy, confidentiality, and compliance risk.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal