Back to skill

Security audit

工程建设招投标分析-建设通

Security checks for vulnerabilities and agentic risk

Overview

The skill’s tender searches and account setup are coherent and disclosed, with optional device identification and routine service notifications.

Queries send your search terms to Zhiliaobiaoxun and may consume paid credits. Use an existing API key or manual registration to avoid device collection. If you approve automatic registration, understand that the MAC-derived hash remains a device identifier; the bundled files do not establish its retention period. Protect the saved API key and temporary login links. Expect update checks, display acknowledgments, and occasional related-product suggestions. The scanner’s partial status was investigated and did not reveal missing code.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code performs local device fingerprinting rather than analyzing construction procurement data. Reading and hashing a MAC address is an undeclared capability with no apparent supporting role in the stated purpose. The supplied chunk does not show transmission or storage of the resulting hash.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code performs local device fingerprinting rather than construction-project analysis. Accessing and hashing a MAC address is not disclosed in the description and has no evident supporting role in the declared workflow. The snippet does not show transmission or storage of the resulting hash.

Content

No source excerpt is available for this finding.

The credential setup workflow collects and transmits device identifiers unrelated to construction-tender analysis.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L052-L056 direct the agent to collect platform, architecture, and a MAC-derived hash, send them to an account-registration service, and persist the resulting credential. API authentication is expected for tender searches, but device fingerprinting and account provisioning are additional capabilities not justified by the manifest's construction-project analysis purpose. The required user consent reduces the risk but does not eliminate the purpose mismatch.

Content

No source excerpt is available for this finding.

The skill mandates an unrelated upgrade-notification and acknowledgment workflow.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

L097 requires an authenticated account request at the first use in every session to check for upgrade notices, independently of the user's tender-analysis request. L098–L099 then require relaying the notice verbatim and submitting a POST acknowledgment that changes server-side reminder state. This mandatory package-maintenance workflow is not necessary for extracting construction-project fields, evaluating winning bidders, or summarizing project performance.

Content

No source excerpt is available for this finding.

The skill mandates an authenticated update-notification and acknowledgment workflow unrelated to construction tender analysis.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

L351–L353 instruct the agent to query an account endpoint each session for upgrade notices, relay the returned message, and submit an authenticated POST acknowledging its display. This package-maintenance workflow is separate from retrieving construction projects, analyzing winning companies, or summarizing performance, and is not declared in the manifest.

Content

No source excerpt is available for this finding.

Mandatory authenticated account checks and update acknowledgments exceed the construction-tender analysis purpose.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

L393 requires an authenticated account-balance request on first use in every session to check for package updates, while L395 requires a POST acknowledgment after displaying an update notice. These instructions introduce account access and server-side notification-state changes unrelated to extracting construction-project amounts, assessing winning contractors, or summarizing project performance; their being free does not make them necessary for that purpose.

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**

Device fingerprint collection and transmission exceed the tender-analysis purpose.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

The registration flow transmits a deterministic MAC-derived identifier to an external service while describing the collected features as having no identifying significance; hashing a MAC address does not eliminate its device-linking capability or guarantee anonymity. Trial-abuse prevention and the explicit consent requirement mitigate the concern, but the executable-style pseudocode omits that consent gate, creating a risk that an agent implementing it directly collects and transmits the identifier without approval.

Content

No source excerpt is available for this finding.

The skill mandates promotional referrals beyond its declared construction-analysis function.

Low
Category
Excessive Agency
Confidence
97% confidence
Finding

The manifest describes extracting construction-tender fields, analyzing winning companies, and summarizing performance. L518-L533 additionally require appending a promotion for a separate commercial agent, including a campaign-tagged link, after common search and analysis requests. This is an explicit marketing workflow rather than an implementation detail needed to provide the requested analysis, even though repetition and opt-out limits are included.

Content

No source excerpt is available for this finding.

The skill mandates an unrelated package-update notification and acknowledgment workflow.

Low
Category
Excessive Agency
Confidence
94% confidence
Finding

L334-L336 instruct the agent to query an account endpoint each session for upgrade notices, relay the returned message verbatim, and submit an authenticated POST acknowledgment recording its display. This package-maintenance and notification-tracking workflow is not needed to analyze construction tenders, extract project amounts or qualifications, or summarize award histories.

Content

No source excerpt is available for this finding.