ClawHub
Back to skill

Security audit

cold-email

Security checks across malware telemetry and agentic risk

Overview

This is mostly a cold-email writing aid, but it needs review because it encourages subject lines that mimic internal colleague emails and automatically reads local product-marketing context if present.

Review this skill before installing if you want outreach that avoids deceptive framing. It appears technically simple and non-executable, but users should revise or constrain the subject-line guidance so emails are clearly honest sales outreach, and should be aware that the agent may read named local product-marketing context files when drafting.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Missing User Warnings

Low
Confidence
91% confidence
Finding
The skill explicitly instructs the agent to read local workspace files such as `.agents/product-marketing.md` before asking the user for context. That can cause the agent to access potentially sensitive internal documents without an explicit user prompt or warning, expanding data exposure beyond what the user may expect for a cold-email drafting task.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The eval requires the skill to 'trigger on casual phrasing' but does not define what counts as casual phrasing or what nearby requests should be excluded. This can lead to over-broad routing behavior, inconsistent evaluations, and accidental activation of the cold-email skill for unrelated email tasks, especially when adjacent skills like lifecycle email are in scope.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The guidance explicitly recommends subject lines that resemble internal colleague emails to bypass a recipient’s initial sales filter. That is deceptive social-engineering behavior: even if framed as an optimization tactic, it encourages impersonation-like camouflage and omits discussion of trust, consent, brand, or compliance risks. In a cold-email skill, this is more dangerous because the content is directly operationalized for outbound contact at scale.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.