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 for vulnerabilities and agentic risk

Overview

The procurement tools fit their purpose, but optional signup understates the privacy implications of its persistent device identifier.

Prefer an existing API key or manual registration to skip device collection. If accepting automatic signup, understand that the provider receives a stable device fingerprint despite the non-identifying wording; retention and deletion practices are unspecified. Searches go to Zhiliaobiaoxun and may consume credits, while update acknowledgements report notice display under your account. Protect the saved API key and temporary login links.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (10)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared purpose concerns procurement-data retrieval and analysis, but the code generates a local device fingerprint. No connection to the declared functionality is established, so this is an undeclared capability rather than an evident supporting implementation detail. The snippet does not show transmission or storage of the identifier.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code derives a machine identifier from local networking hardware rather than querying or analyzing procurement information. Hardware fingerprinting is not disclosed in the declared purpose, and the chunk provides no evidence that it is necessary to support the advertised functionality. No transmission or persistence of the resulting hash is shown.

Content

No source excerpt is available for this finding.

The skill collects and transmits device fingerprints for account registration, beyond procurement-data querying and analysis.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

L053–L055 instruct the agent to collect platform, architecture, and a hashed MAC address and submit them to an automatic-registration endpoint. Authenticating procurement API requests is expected, but fingerprinting the user's device to provision an account is a separate capability not justified by the manifest's analytics purpose. Requiring user consent mitigates the privacy risk but does not make device fingerprinting necessary for the stated task.

Content

No source excerpt is available for this finding.

The query workflow lacks a user-facing disclosure that user-provided search inputs are transmitted to a third-party service.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The execution workflow uses user-provided keywords, company names, and filters in requests to the external Zhiliaobiaoxun API documented at L019–L034, without instructing the agent to disclose that transmission to the user. The consent requirement at L052–L054 covers device characteristics during registration, not routine query data that may reveal business interests.

Content

No source excerpt is available for this finding.

Collects and transmits a persistent hardware fingerprint for trial-account deduplication, beyond procurement search and analysis.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The workflow sends a deterministic, unsalted hash of a normalized MAC address while describing the collected features as having no identity meaning and collecting no personal information. Hashing does not remove the identifier's persistent linkability, so these assurances can undermine informed consent and expose users to device correlation, although the explicit consent gate and manual alternative reduce the risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auto-register.md (reported line 130)May include surrounding context.

md
> ### ⚠️ 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`)。
>

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auto-register.md (reported line 131)May include surrounding context.

md
> ### ⚠️ 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`)。
>
> 历史教训:曾有版本采集 `home_path`,Windows 的 `C:\Users\alice` 直接拼进 JSON 字符串时

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/auto-register.md (reported line 195)May include surrounding context.

text

注意事项:
- 目录不存在时先 `mkdir -p ~/.zlbx`
- 文件已存在时**合并而非覆盖**(保留用户可能的其他配置)
- `source: "auto"` 字段必须写入,**这是后续判断「是否输出自动登录链接」的关键依据**

The skill mandates a remote upgrade-notice acknowledgement workflow unrelated to procurement research and analysis.

Low
Category
Excessive Agency
Confidence
90% confidence
Finding

The instructions require a session-start upgrade check, forwarding upgrade notices to the user, and an authenticated POST confirming that a notice was displayed. That POST updates server-side notification state rather than retrieving or analyzing procurement data, and the manifest does not declare upgrade-notification management as a capability.

Content

No source excerpt is available for this finding.

The update-notice workflow sends an account-linked display acknowledgement without a user-facing privacy disclosure.

Low
Category
Output Handling
Confidence
88% confidence
Finding

The skill instructs the agent to POST a notice ID and displayed: true using the user's existing credentials after showing an update notice. This reports the user's notice-display status to the service, but the workflow provides no user-facing disclosure of that reporting and explicitly keeps the technical acknowledgement out of the conversation.

Content

No source excerpt is available for this finding.