ClawHub
Back to skill

Security audit

cs

Security checks across malware telemetry and agentic risk

Overview

This is a simple summarization skill with disclosed automatic triggers; its main risk is that long messages may be sent to the configured LLM for summarization.

Install this only if you want automatic summaries of long messages. Be aware that messages over 100 characters can trigger without an explicit summarize request, and the message content will be processed by your configured LLM provider.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Confidence
92% confidence
Finding
Automatically activating summarization for any plain-text message over 100 characters is an overly broad trigger that can cause the skill to process content without clear user intent. In an agent setting, this can lead to unintended handling of sensitive emails, chats, or notes, and may alter normal workflow by injecting summaries when the user did not request them.

Vague Triggers

Medium
Confidence
88% confidence
Finding
Using broad everyday trigger terms like 'brief' increases the chance of accidental activation during normal conversation, especially in multilingual or casual exchanges. While less severe than the length-based trigger, it still creates ambiguity around user intent and can cause unnecessary or privacy-impacting processing of message content.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill auto-matches not only on explicit summary requests but also on any message longer than 100 characters, which can cause silent invocation on unrelated or sensitive user content without clear disclosure or consent. In an agent setting, this increases the chance that private text is unexpectedly forwarded to the LLM for processing, creating privacy and transparency risks.

Natural-Language Policy Violations

Medium
Confidence
87% confidence
Finding
The skill hardcodes Chinese-language behavior in its name, description, prompt, and system message without checking the user's preferred language. This is primarily a safety and usability issue: it can produce unexpected output language, confuse users, and degrade reliability, though it does not by itself create a direct security exploit.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.