ClawHub
Back to skill

Security audit

XHS

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly coherent for Xiaohongshu browser automation, but one draft-saving instruction appears to target the publish button, creating a real risk of unintended public posting.

Review this skill before installing if you use it on a real Xiaohongshu account. It should be corrected so draft-saving clicks the draft/save control, not the red publish control, and you should require explicit confirmation immediately before any public post or account reply.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Context-Inappropriate Capability

Medium
Confidence
89% confidence
Finding
The manifest describes a browser-automation skill for drafting, publishing, replying, and checking metrics in Xiaohongshu Creator Center. Line L123 instructs the agent to run a local `cp` command to move files between directories, which is a host filesystem/shell capability not justified by the stated purpose and goes beyond ordinary browser automation.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The README says '直接描述你的目标即可' ('just describe your goal directly'), which does not define clear trigger boundaries or exclusions. This broad activation guidance overlaps with ordinary conversation and lacks negative examples or scope constraints, making it ambiguous when the skill should activate.

VirusTotal

48/48 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.