Back to skill

Security audit

医疗大健康采招雷达-医疗招标采购网

Security checks for vulnerabilities and agentic risk

Overview

This skill needs review because signup understates device identification and conflicting recovery instructions could lead to an unintended account change.

Ask the publisher to reconcile credential-recovery instructions and clarify device-identifier privacy before installing. Prefer manual registration with your own API key, and require explicit approval before replacing existing credentials. Search terms go to Zhiliaobiaoxun and queries can consume credits. Protect the saved key and generated login links. All bundled files were inspected; the service's actual data-retention practices remain unverified.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code accesses local network-interface metadata to generate a device fingerprint, rather than retrieving or analyzing medical procurement information. This capability is not disclosed or evidently necessary for the declared purpose. The supplied chunk does not transmit the identifier or perform procurement analysis.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code performs local device fingerprinting rather than healthcare procurement research. Accessing and hashing a MAC address is not an evident supporting implementation detail of the declared search and supplier-analysis purpose. No procurement search, extraction, or analysis appears in this chunk, and no transmission of the resulting hash is shown.

Content

No source excerpt is available for this finding.

Activation is based on broad healthcare keywords rather than procurement intent.

Medium
Category
Trigger Abuse
Confidence
95% confidence
Finding

The description activates the skill whenever search terms contain 医院 (hospital), 医疗 (medical), 卫生 (health/hygiene), or 体检 (physical examination). These terms also occur in ordinary healthcare searches, so the condition can invoke a procurement-data skill without any bidding or purchasing intent; the later routing exclusions do not exclude these non-procurement requests.

Content

No source excerpt is available for this finding.

API onboarding collects and transmits device fingerprints beyond the medical procurement analysis purpose.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L053-L055 instruct the agent to collect platform, architecture, and a MAC-derived hash and submit them to an account-registration endpoint. Device fingerprinting is not necessary to search medical procurement notices or analyze hospital suppliers. Explicit user consent reduces the privacy risk but does not make this capability part of the manifest's stated analytical purpose.

Content

No source excerpt is available for this finding.

Authentication-error handling contradicts the restriction that automatic registration occurs only when no key is configured.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The error table directs automatic registration even when an existing key is invalid, expired, or disabled, contradicting the earlier restriction that registration occurs only when no key is configured. An authentication failure could therefore trigger unintended registration, device-data collection if consent is subsequently granted, or replacement of locally stored credentials rather than recovery of the user's intended account; the general consent requirement reduces but does not resolve this account-integrity risk.

Content

No source excerpt is available for this finding.

The skill mandates a per-session upgrade-notification and acknowledgement workflow beyond medical procurement analysis.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The manifest describes medical procurement searches, buyer/winner extraction, and hospital supplier analysis. These instructions additionally require an account request on first use each session, forwarding server-provided upgrade notices, and submitting a POST acknowledgement after display, even when the user has requested only procurement analysis.

Content

No source excerpt is available for this finding.

Automatic company searches transmit user-provided company names to an external API without a privacy disclosure.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The workflow instructs the agent to send the user's company name to search_company and automatically use the matched company list in subsequent queries without interrupting the user. Unlike a passive API catalog entry, this is an instruction to transmit user-supplied information, but the file does not warn users that their company queries will be shared with the external 知了标讯 service.

Content

No source excerpt is available for this finding.

The skill defines a standalone corporate-shareholder lookup workflow outside its declared healthcare procurement scope.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

The manifest describes healthcare tender searches, extracting hospitals and winning suppliers, and analyzing a hospital’s top suppliers. L124–L132 instead prescribe shareholder reporting and explicitly direct the agent to handle shareholder-only requests using get_company_shareholders, without a healthcare procurement connection; this is an independent corporate-ownership service rather than an implementation detail of the declared workflow.

Content

No source excerpt is available for this finding.

L080–L119 collect a hardware-derived device fingerprint and transmit it for account registration, beyond the stated medical procurement analysis purpose.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The registration workflow transmits a stable hardware-derived identifier while describing the collected features as having no identity implications and collecting no personal information. An unsalted MAC-address hash is a linkable pseudonymous identifier, not guaranteed anonymization, so this wording understates the privacy consequences and weakens informed consent despite the explicit approval gate.

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**

Mandatory account-based update tracking is unrelated to medical procurement analysis.

Low
Category
Excessive Agency
Confidence
91% confidence
Finding

L351 requires an authenticated account-balance request at the first use of every session specifically to check package-update notices, rather than to perform procurement analysis. L352–L353 additionally require relaying the notice and submitting an authenticated display acknowledgment, introducing an update-tracking workflow outside the manifest's hospital and supplier-analysis purpose.

Content

No source excerpt is available for this finding.

L393-L395 mandate an upgrade-notification and remote acknowledgment workflow unrelated to medical procurement analysis.

Low
Category
Excessive Agency
Confidence
90% confidence
Finding

The skill instructs the agent to query an account endpoint at first use to check upgrade notices, relay server-provided messages, and submit an authenticated POST acknowledgment after displaying them. This adds package-update notification delivery and remote acknowledgment state changes, neither of which is needed to identify hospital purchasers, winning suppliers, or supplier rankings.

Content

No source excerpt is available for this finding.

The skill mandates an authenticated account/update-notification workflow unrelated to medical procurement analysis.

Low
Category
Excessive Agency
Confidence
93% confidence
Finding

L334 requires an authenticated account-balance request each session specifically to check upgrade notices, while L335-L336 require relaying those notices and submitting a POST acknowledgement that changes server-side notification state. These are package-maintenance and account operations, not necessary steps for retrieving medical procurement notices or analyzing hospital suppliers.

Content

No source excerpt is available for this finding.