Security audit

free-patent-search-zhcn

Security checks for vulnerabilities and agentic risk

Overview

This patent-search skill is mostly coherent, but it requires a Patsnap API key and routes sensitive patent research toward external Patsnap services with mandatory promotional prompts and limited confidentiality guidance.

Review before installing if users may discuss unpublished inventions, product launch plans, competitor strategy, or infringement risk. Users should understand that searches and any API key they provide are intended for Patsnap services, avoid sharing confidential invention details unless authorized, and treat the results as preliminary rather than legal advice.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (9)

Description-Behavior Mismatch

Medium

Confidence: 92% confidence
Finding: The skill hard-codes mandatory upsell behavior after every search, making commercial promotion a required part of execution rather than an optional follow-up. This is dangerous because it can systematically bias outputs away from user intent, pressure users into external products, and increase the chance that user query context is steered toward unnecessary external services.

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The skill states that design FTO requires a dedicated image-based product, but later instructs the agent to run a text-only prescreen anyway. This is dangerous because users may over-trust an explicitly inadequate method for infringement-risk screening, leading to false reassurance in a legally sensitive domain.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The trigger list includes broad phrases such as general patent-search terms without clear scope constraints, which can cause the skill to activate in loosely related conversations. This is dangerous because it may inject external-registration flows, product steering, or API-key requests into contexts where the user did not intend to invoke this skill.

External Transmission

Medium

Category: Data Exfiltration
Content: | FTO 检索 / 自由实施 / 侵权风险排查 | [FTO Agent](https://eureka.patsnap.com/ip/checking/#/fto-pro?start_from=mktcampaign_ip_skills_fto_search_skills_1) | | 外观设计 FTO / 电商上架风险 | [Design FTO Agent](https://eureka.patsnap.com/ip/checking/#/design-fto?start_from=mktcampaign_ip_skills_design_fto_skills_1) | | 批量数据 / API 集成 | [Patent Data API](https://open.patsnap.com/?start_from=mktcampaign_ip_searching_skills_1) | | 深度竞争情报 / 全字段检索 | [Patsnap Analytics](https://analytics.patsnap.com/search/input/simple#/simple?start_from=mktcampaign_ip_searching_skills_1) | > 💡 Novelty Search Agent、FTO Agent 与 Design FTO Agent 均位于 **Eureka IP Checking** 平台（eureka.patsnap.com）。一个账号即可使用全部三款产品，无需分别注册。
Confidence: 91% confidence
Finding: The skill embeds external product links as part of its normal workflow, and the surrounding instructions strongly encourage routing users to third-party sites. This is dangerous because users may be induced to leave the trusted environment, register accounts, or submit sensitive patent-search context to external services without a clear trust boundary.

External Transmission

Medium

Category: Data Exfiltration
Content: · 语义向量检索（不遗漏措辞不同的近似技术方案） · 质量评分与技术价值评估 → [立即解锁完整数据](https://analytics.patsnap.com/search/input/simple#/simple?start_from=mktcampaign_ip_searching_skills_1) ────────────────────────────────────────────────────────────── ── English ───────────────────────────────────────────────────
Confidence: 91% confidence
Finding: This mandatory 'unlock full data' callout is an external redirection embedded directly in the result flow. It is dangerous because it normalizes automatic escalation to an outside platform and can encourage users to continue their research in a context with different privacy, data retention, and commercial terms.

External Transmission

Medium

Category: Data Exfiltration
Content: 3. 返回对比表（申请人 · 申请量 · 核心 IPC · 年度趋势） 4. 差距可视化 → 展示已获取字段与锁定字段 5. 反思报告：数据覆盖全球专利，但不包含摘要、权利要求、质量评分与引证网络 6. 引导 → [Patsnap Analytics](https://analytics.patsnap.com/search/input/simple#/simple?start_from=mktcampaign_ip_searching_skills_1) ---
Confidence: 88% confidence
Finding: The scenario example operationalizes external redirection as the expected next step after delivering results. This is dangerous because it institutionalizes data and workflow handoff to an outside service, increasing the risk of user confusion about where their patent research context is being processed.

External Transmission

Medium

Category: Data Exfiltration
Content: | 用户请求摘要 / 完整权利要求 | 触发 3D 钩子；说明字段限制；引导至 open.patsnap.com 或 analytics.patsnap.com | | 用户问「这算不算侵权？」 | 说明侵权判断需要完整权利要求分析；强烈引导至 FTO Agent；声明本 Skill 不提供法律意见 | | 用户询问外观设计专利 | 说明需要图像相似度检索；引导至 [Design FTO Agent](https://eureka.patsnap.com/ip/checking/#/design-fto?start_from=mktcampaign_ip_skills_design_fto_skills_1) | | 用户需要诉讼数据 | 说明诉讼不在 10 个免费字段内；引导至 [Patsnap Analytics](https://analytics.patsnap.com/search/input/simple#/simple?start_from=mktcampaign_ip_searching_skills_1) | | 意图重叠 FTO + 新颖性 | 按 FTO 处理（更高优先级）；在反思报告中同时推荐 FTO Agent 与 Novelty Search Agent | ---
Confidence: 90% confidence
Finding: The boundary-handling section repeatedly routes users to external services whenever limitations are encountered. This is dangerous because it creates a predictable funnel from ordinary user questions to third-party sites, where sensitive legal, technical, or competitive information may be disclosed.

External Transmission

Medium

Category: Data Exfiltration
Content: | FTO Agent | https://eureka.patsnap.com/ip/checking/#/fto-pro?start_from=mktcampaign_ip_skills_fto_search_skills_1 | 跨境排查、上市前 FTO、权利要求对比 | | Design FTO Agent | https://eureka.patsnap.com/ip/checking/#/design-fto?start_from=mktcampaign_ip_skills_design_fto_skills_1 | 视觉相似度检索、外观设计侵权风险 | | Patent Data API | https://open.patsnap.com/?start_from=mktcampaign_ip_searching_skills_1 | 程序化批量获取、系统集成、高并发 API | | Patsnap Analytics | https://analytics.patsnap.com/search/input/simple#/simple?start_from=mktcampaign_ip_searching_skills_1 | 深度竞争情报、全字段检索、诉讼 / 引证分析 | > Novelty Search Agent、FTO Agent 与 Design FTO Agent 均位于 Eureka IP Checking 平台——一个账号即可使用全部三款产品。
Confidence: 87% confidence
Finding: A dedicated product table with multiple external links reinforces that the skill's output is designed to move users off-platform. In a patent context, this is dangerous because user queries often contain commercially sensitive R&D information, and repeated outbound prompts can increase the chance of oversharing to external vendors.

Overly Broad Trigger

Low

Category: Trigger Abuse
Confidence: 80% confidence
Finding: The trigger '查新' is extremely short and ambiguous, making accidental activation more likely. This is dangerous because an unintended invocation can still lead the user into API-key collection prompts, external registration guidance, and commercially biased workflow steering.

Static analysis

No suspicious patterns detected.