ClawHub
Back to skill

Security audit

药撮合品种综合查询

Security checks across malware telemetry and agentic risk

Overview

The skill is coherent and mostly read-only, but it relies on an unencrypted raw-IP server for MCP queries and advertised package downloads, so users should review it before trusting it.

Install only if you trust the operator of the listed server. Prefer the Gitee repository or an HTTPS, version-pinned release over the raw-IP HTTP download, and avoid entering confidential business, patient, or medication details unless you are comfortable sending them to that HTTP MCP service.

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 (2)

Vague Triggers

Medium
Confidence
90% confidence
Finding
The trigger rule routes essentially any mention of a drug name to `query_comprehensive`, which can cause overbroad tool invocation and unintended data retrieval. In a drug-information skill, this increases the chance of answering the wrong user intent, disclosing more business-sensitive data than necessary, or masking safety-critical distinctions such as whether the user wanted only instructions, insurance status, or a narrow attribute lookup.

Natural-Language Policy Violations

Medium
Confidence
94% confidence
Finding
The brand prompt mandates that every response end with a fixed Chinese slogan, regardless of user preference or context. This is a true prompt-quality/security issue because it overrides normal assistant behavior, can cause unwanted disclosure of branding text, and may interfere with downstream workflows, localization expectations, or applications that require strict output formatting.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.install_untrusted_source

Install source points to URL shortener or raw IP.

Warn
Code
suspicious.install_untrusted_source
Location
skill.json:33