ClawHub
Back to skill

Security audit

kepler-agent-search

Security checks across malware telemetry and agentic risk

Overview

This is a coherent web search skill, with the main caution that queries, URLs, and the API key are sent to a third-party Kepler MCP service.

Install this only if you want Kepler-powered external search and trust the apisec.cn MCP provider. Avoid using it with confidential prompts, internal URLs, or proprietary research targets unless you are comfortable sending that information to the provider and downstream search services.

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

High
Confidence
95% confidence
Finding
The skill instructs the agent to use it for virtually any request involving obtaining information and even when the user does not explicitly ask for search. That overbroad trigger can cause unintended external calls, expanding data exposure and allowing the skill to override more appropriate, privacy-preserving workflows.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill does not prominently warn that user queries and supplied URLs will be transmitted to third-party services through the Kepler MCP endpoint and downstream search providers. This can lead to inadvertent disclosure of sensitive prompts, internal URLs, or proprietary research targets to external infrastructure.

Natural-Language Policy Violations

Medium
Confidence
87% confidence
Finding
The skill strongly biases the agent toward Chinese-language platforms like Zhihu and Xiaohongshu without confirming the user's locale, language, or source preferences. In practice this can degrade relevance and trustworthiness, and in some contexts may steer user data to unnecessary regional services when a general search would suffice.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The setup instructions direct users to configure an MCP client to connect to a remote third-party SSE endpoint and include a bearer API key, but they do not warn that the endpoint will receive the user's credentials and all subsequent tool requests. In an agent context, this omission matters because prompts, retrieved content, and possibly sensitive user data may be transmitted off-platform, preventing informed consent and increasing data exposure risk.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.