Lobster Attachment Inject

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:35 PM UTC
• Version:
  1.0.1

Overview

This is a markdown-only helper for reading local agent and skill descriptions and optionally generating lightweight registry text, with no hidden code execution or data export found.

Install only if you are comfortable with the skill reading local agent and skill markdown files and helping create or update registry/prompt-context files. Review any generated AGENTS.md or registry changes before relying on them, especially in repositories containing private workflows or sensitive notes.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The activation examples use broad natural-language phrases like "调度Agent" and "调试" as triggers without defining scope, precedence, or disambiguation rules. In an agentic system, vague trigger boundaries can cause unintended skill or file loading, which increases the chance of prompt-routing mistakes, unnecessary exposure of internal registries, or attacker-controlled phrasing steering the agent into unexpected behavior.

VirusTotal

52/52 vendors flagged this skill as clean.