Back to skill

Security audit

LobsterMail

Security checks across malware telemetry and agentic risk

Overview

The skill provides the email-agent features it advertises, but its install and update instructions give a remote service too much ongoing control over local agent code.

Install only if you trust LobsterMail with email contents and metadata and are comfortable with a local reusable token. Avoid the pipe-to-shell installer, inspect any downloaded installer or package first, and do not allow daily self-updates unless each update is explicitly reviewed. Avoid using this for MFA, password resets, regulated data, production accounts, or outbound email unless the user has clearly approved the exact workflow.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
The manifest advertises 'No API keys, no human signup,' but the documented REST flow later requires account creation and bearer-token authentication. This mismatch can mislead operators and agents into treating the integration as lower risk than it is, causing unreviewed credential creation and use of an external service.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The body text repeats the claim that no API keys or human signup are needed, yet later instructions describe explicit signup and token-based authorization. Security-relevant capability mismatches reduce informed consent and may cause agents to handle tokens and external communications without appropriate safeguards.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
The README promotes autonomous creation of email inboxes, sending/receiving real email, and local token persistence, but does not clearly warn users about privacy, abuse, compliance, and system-impact implications. In an agent skill context, this is materially risky because it enables external communications and account creation that could be used for spam, data exfiltration, or interacting with third-party services without informed operator consent.

Missing User Warnings

High
Confidence
98% confidence
Finding
The installation instructions recommend piping a remotely fetched script directly into a shell, which executes unreviewed code from the network immediately. This is dangerous because any compromise of the remote endpoint, CDN, TLS termination, or install script can lead to arbitrary code execution on the host running the agent skill.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The README states that a token is persisted to ~/.lobstermail/token without warning about credential storage, file permissions, rotation, or host compromise implications. In an agent environment, a locally stored reusable token can be harvested by other processes or users and abused to access or operate the email account.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill enables real email send/receive through a third-party service but does not prominently warn that message bodies, metadata, addresses, and potentially sensitive verification emails are transmitted to and stored by that service. In this context, lack of disclosure is risky because users may provide confidential data assuming local-only handling.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The SDK auto-signs up and persists a token to ~/.lobstermail/token without a strong user-facing warning. Automatic credential creation and local persistence expand the attack surface, especially on shared systems or in agent environments where filesystem secrets may be exposed to other tools or users.

External Script Fetching

High
Category
Supply Chain
Content
One-liner — downloads the skill from the edge proxy and saves it to your OpenClaw skills directory:

```bash
curl -sL https://api.lobstermail.ai/install | sh
```

Restart OpenClaw (or start a new session). The skill loads automatically.
Confidence
99% confidence
Finding
curl -sL https://api.lobstermail.ai/install | sh

Chaining Abuse

High
Category
Tool Misuse
Content
One-liner — downloads the skill from the edge proxy and saves it to your OpenClaw skills directory:

```bash
curl -sL https://api.lobstermail.ai/install | sh
```

Restart OpenClaw (or start a new session). The skill loads automatically.
Confidence
97% confidence
Finding
| sh

VirusTotal

No VirusTotal findings

View on VirusTotal