Security audit

邮件日报免费版

Security checks across malware telemetry and agentic risk

Overview

The skill does what it claims, but it asks agents to access email accounts and includes a password-entry command pattern that could expose credentials in logs.

Review before installing. Use only with accounts you are comfortable exposing to the agent and browser automation environment. Prefer an already logged-in browser session, do not type real passwords into command examples or prompts, and delete or protect generated screenshots because they may contain private email content.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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)

Intent-Code Divergence

High

Confidence: 97% confidence
Finding: The skill claims it avoids storing passwords in scripts, yet elsewhere provides a manual login example with a plaintext password literal. This creates a real credential-exposure risk because users or agents may place secrets directly into commands, shell history, logs, transcripts, or skill artifacts.

Intent-Code Divergence

High

Confidence: 99% confidence
Finding: The '安全优先原则' recommends environment variables for passwords, but the manual login workflow contradicts that guidance by showing direct plaintext password entry. In an agent context with exec/browser tooling, this is especially dangerous because command transcripts and tool logs can persist sensitive credentials.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill instructs users to capture mailbox screenshots and extract message content, which can expose sensitive personal or business information. In an agent environment, these artifacts may be retained, shared, or inspected later, so the absence of strong privacy warnings and retention limits creates meaningful data-handling risk.

Missing User Warnings

High

Confidence: 98% confidence
Finding: The manual login example tells the operator to enter a password directly via command-driven browser automation without warning that credentials may appear in shell history, agent logs, tool telemetry, or conversation transcripts. Because this skill targets email accounts, compromise of those credentials could expose a large volume of sensitive communications and enable account takeover.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.