ClawHub
Back to skill

Security audit

Xiaohongshu User Profile

Security checks across malware telemetry and agentic risk

Overview

This skill mainly does what it says, but it also gives guidance for scaled Xiaohongshu scraping with stealth browser sessions and under-discloses local note persistence.

Review before installing. Use only for user-directed, limited lookups where you are comfortable logging into Xiaohongshu in the agent-controlled browser. Avoid using the batch and stealth multi-session guidance unless you have clear authorization and understand platform rules. Check or disable the local memory-file behavior if you do not want operational notes written into the working directory.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The skill goes beyond a single user-profile lookup and explicitly instructs batch orchestration, adding intervals to avoid rate limits, and using multiple stealth browser sessions with independent fingerprints. That materially increases scraping scale and provides operational guidance for evading platform anti-abuse controls, which raises the risk of unauthorized mass collection and account/session abuse.

Intent-Code Divergence

Medium
Confidence
83% confidence
Finding
The skill claims its boundary is limited to reading data visible in the browser, but elsewhere it instructs reading and appending a local memory file. That mismatch expands the skill's effective access from browser-only extraction to local filesystem interaction, which can surprise operators and create unintended persistence or leakage of execution metadata.

Description-Behavior Mismatch

Low
Confidence
76% confidence
Finding
The manifest describes profile and notes retrieval, but the skill also performs local experience-file reads and writes that are not part of the stated purpose. Even if the stored content is operational rather than user data, hidden persistence broadens the skill's behavior and can create auditability, privacy, and trust issues.

Missing User Warnings

Low
Confidence
80% confidence
Finding
The skill instructs writing to a memory file without a clear user-facing warning at the point of use. Silent persistence is risky because users may not expect local state to be modified, and accumulated notes about failures or strategy changes can still reveal environment details or operational history.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.