ClawHub
Back to skill

Security audit

医药企业综合情报（Pharma Company Intelligence）

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed pharma intelligence workflow that searches public and patent sources, optionally reads a user-provided reference document, and writes a session HTML report.

Install only if you want a pharma/biotech diligence workflow that can run extensive web and patent searches and create a session report. Do not provide confidential PDFs, PPTs, or DOCX files unless you intend the skill to read and summarize them into the report.

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

Medium
Confidence
94% confidence
Finding
The trigger phrases are broad, natural-language requests like '帮我做一份 [公司名] 的调研报告' and '我想了解 [公司名] 的药物管线和专利情况', which overlap with ordinary user queries. This increases the chance of unintended activation, causing the skill to automatically read optional uploaded documents and launch extensive web/patent research workflows when the user may not have explicitly chosen this skill.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill accepts a `reference_doc` input and later states it will read the uploaded file and generate a persistent HTML report under `@session/reports/`, but the description does not clearly warn users about these behaviors up front. This can undermine informed consent by surprising users that their uploaded documents will be parsed and their data written to a report file, potentially exposing sensitive business or diligence material.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The manifest describes a very broad, multi-module intelligence capability without clear activation boundaries, inputs, or guardrails. This increases the risk that the skill is invoked for loosely related requests and performs expansive data gathering or analysis beyond user intent, which is especially sensitive given the external search and patent-analysis dependencies.

Natural-Language Policy Violations

Medium
Confidence
82% confidence
Finding
The manifest text specifies Chinese-language output ('HTML报告' and Chinese-market reporting) without indicating that language is user-selectable or justified by context. Hard-coding output language can mislead orchestration, reduce transparency for non-Chinese users, and create opportunities for hidden or less-reviewable output in multilingual environments.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.