Back to skill

Security audit

政府采购招标大数据分析-政采商机与中标洞察

Security checks across malware telemetry and agentic risk

Overview

The skill provides procurement analytics, but it also provisions and stores an account using device-derived identifiers and generates billing login links, so users should review it before installing.

Install only if you are comfortable using the zhiliaobiaoxun external service for procurement queries and with its onboarding flow. To avoid automatic registration, configure your own ZLBX_API_KEY or ~/.zlbx/config.json first; if no key is present, require explicit consent before device-feature collection, account creation, or credential storage.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (11)

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
The skill documentation goes beyond procurement analysis and instructs the agent to automatically register an account, collect device-derived identifiers, and persist a returned API key locally. That expands the skill into account lifecycle and credential management, creating privacy and unauthorized-account-creation risks that are not necessary for the requested analysis task.

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
The skill instructs generation of an auto-login recharge link tied to a device/session identifier, which extends the agent into authentication and account-access workflows unrelated to core procurement analysis. This can enable unintended account access transitions and encourages the agent to broker login flows on the user's behalf.

Description-Behavior Mismatch

High
Confidence
97% confidence
Finding
The document defines a full auto-registration, account provisioning, persistence, and recharge lifecycle that goes well beyond the skill’s declared procurement-analysis purpose. This creates hidden account-management behavior, including device-derived data collection and outbound registration requests, which materially expands the trust boundary and can surprise users who only expected data analysis functionality.

Context-Inappropriate Capability

High
Confidence
98% confidence
Finding
The skill instructs collection of device fingerprint attributes, including architecture and a hashed MAC-derived identifier, and sends them to a remote service for deduplication. Even if the raw MAC is not transmitted, this is still a persistent device identifier unrelated to procurement analysis and enables user/device tracking beyond the stated function of the skill.

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
The skill persists API keys into a local config file, creating a lasting credential on disk outside the manifest’s analytical purpose. Local persistence increases exposure to credential theft, accidental reuse across contexts, and silent future authentication behavior that the user may not expect.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The quota-exhaustion flow generates a device SID and directs the user into an auto-login and phone-binding/recharge path, which is unrelated to procurement analytics. Embedding this commercial account-upgrade workflow in the skill increases the chance of unanticipated external interactions and expands data processing beyond the declared purpose.

Vague Triggers

High
Confidence
90% confidence
Finding
The activation rule is overly broad and says the skill must be used whenever common government/procurement-related keywords appear, even if the user did not explicitly request this capability. Overbroad mandatory invocation can cause unnecessary data sharing with the external service and increases the chance the agent routes unrelated or sensitive requests into this skill.

Missing User Warnings

High
Confidence
98% confidence
Finding
These instructions direct silent collection of device characteristics, a network registration request, and writing credentials to a local file, all without explicit notice or consent. That combines privacy-sensitive fingerprinting with persistent local state changes, creating significant risk of unauthorized system modification and secret handling by the agent.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill explicitly instructs the agent not to disclose automatic registration, login, or device-binding behavior in certain cases. Suppressing transparency about credential provenance and account-binding actions undermines informed consent and makes potentially sensitive side effects harder for users to detect.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The documentation explicitly tells the agent to automatically select all semantically matching companies, including headquarters and subsidiaries, and proceed without user confirmation even when the input is only a shorthand name. In this procurement-analysis context, that can cause entity confusion, over-collection of unrelated company data, and materially incorrect analysis or outreach decisions based on the wrong legal entities.

External Transmission

Medium
Category
Data Exfiltration
Content
> ### ⚠️ 请求体必须用 JSON 序列化函数生成,不要手拼字符串
>
> 用 `json.dumps(payload)` / `requests.post(url, json=payload)` / `JSON.stringify(payload)`,
> 或 `curl -d @file`;**不要用字符串拼接,也不要用 Python 的 `str(dict)`**
> (后者产出单引号,服务端会报 `Expecting property name enclosed in double quotes`)。
>
Confidence
93% confidence
Finding
requests.post(url, json=

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.