ClawHub

血友病公众号文章写作

Security checks across malware telemetry and agentic risk

Overview

This WeChat article-writing skill is mostly coherent, but it automatically tells the agent to read the user's local CLAUDE.md file without a consent step or scope limits.

Install only if you are comfortable with the agent reading CLAUDE.md during article-writing requests. Safer use would be to edit the skill to ask first, narrow activation to explicit WeChat/公众号 article tasks, or provide writing-style preferences directly instead of reading a local file.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
85% confidence
Finding
The trigger description is broad enough to activate on generic content-writing requests, which can cause the skill to run outside its intended WeChat-specific scope. Overbroad activation increases the chance that users are funneled into behaviors like web searching or local file access they did not explicitly request, creating consent and data-exposure risks.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The instruction to read the user's CLAUDE.md directs the agent to access a local file without an explicit user-facing permission step or disclosure of what data will be read. Because CLAUDE.md may contain private preferences, project instructions, secrets, or other sensitive local context, this can lead to unintended disclosure and use of data beyond what the user asked for.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal