Back to skill

Security audit

施工建材采招助手-鲁班乐标

Security checks for vulnerabilities and agentic risk

Overview

The search features fit the stated purpose, but device-identification disclosures and account-key safeguards need review.

Prefer supplying your own API key to skip hardware-based registration. Otherwise, understand that the MAC hash can consistently identify a device despite the privacy wording. Queries send research terms to Zhiliaobiaoxun and may consume paid credits. Keep saved keys and automatic-login links private. The publisher should clarify acknowledgment destinations and consistently include consent in registration examples. All bundled files and reported reference gaps were reviewed; server-side retention and acknowledgment behavior 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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (14)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code performs local device fingerprinting rather than construction-material price or supplier queries. Reading and hashing a MAC address is not an evident supporting implementation detail of the declared procurement functionality. The chunk contains no price-trend or top-brand API calls and shows no transmission of the generated 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 hardware identification rather than price-trend or top-brand queries. Reading and hashing a MAC address is not an evident supporting implementation detail of the declared procurement function. The chunk shows no transmission of the resulting identifier.

Content

No source excerpt is available for this finding.

Automatic account enrollment collects device identifiers and persists credentials outside the stated material-query purpose.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

L052-L056 instruct the agent to collect platform, architecture, and a MAC-derived hash, transmit them to an account-registration endpoint, and write the resulting credential to local configuration. Using an API key is expected for remote material queries, but device fingerprinting and account provisioning are distinct capabilities not declared in L003. The required user consent mitigates the privacy concern but does not make these capabilities part of the stated purpose.

Content

No source excerpt is available for this finding.

The skill exposes general corporate-intelligence capabilities beyond construction-material pricing and supplier identification.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The manifest at L003 limits the skill to construction-material historical prices and major suppliers. L077-L084 instead expose shareholder records, investments, corporate changes, project contacts, and competitor analysis; L342-L360 also prescribe a general company-analysis workflow. These are broader intelligence operations, not necessary implementation details of the declared price-and-supplier lookup.

Content

No source excerpt is available for this finding.

The skill mandates an upgrade-notification and acknowledgment workflow beyond its declared material-price and supplier lookup purpose.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest describes construction-material price trends and leading suppliers, but L097 requires an account-balance request at the first use of each session specifically to check upgrade notices. L098–L099 additionally require relaying the notice verbatim and submitting a POST acknowledgment, introducing user-facing upgrade messaging and a remote state change unrelated to the declared lookup outputs.

Content

No source excerpt is available for this finding.

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

Medium
Category
Output Handling
Confidence
86% confidence
Finding

The workflow instructs the agent to send the company named by the user to search_company automatically and use the results for subsequent queries. Unlike a passive API catalog entry, this selects an external operation, but the file does not instruct the agent to disclose that the user's company-search terms will be shared with the 知了标讯 service.

Content

No source excerpt is available for this finding.

The skill mandates a package-update notification and acknowledgment workflow unrelated to material pricing or supplier discovery.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L351 requires an account request each session specifically to check upgrade notices, and L352–L353 require displaying those notices and submitting an authenticated POST acknowledgment. This package-maintenance workflow, including changing server-side acknowledgment state, is not needed to retrieve historical material prices or identify major suppliers and is not declared in the manifest.

Content

No source excerpt is available for this finding.

L393-L395 mandate authenticated account and update-acknowledgment operations unrelated to construction-material sourcing.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest describes retrieving construction-material prices, top brands, and suppliers. L393 instead requires a session-start account-balance request specifically to check upgrade notices, and L395 conditionally requires an authenticated POST recording that a notice was displayed; neither operation is necessary for the declared sourcing workflow.

Content

No source excerpt is available for this finding.

The registry exposes corporate ownership and financial-report queries beyond the material-pricing and supplier-listing purpose.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

The manifest limits the skill to construction-material price trends and leading brands/suppliers. L048–L063 document shareholder, investment, and corporate-change queries, while L068–L078 provide annual-report access covering financial results, social insurance, equity changes, and guarantees. These paid corporate-investigation capabilities are not necessary to retrieve material prices or identify major suppliers and are not declared in the manifest.

Content

No source excerpt is available for this finding.

L334-L336 require an authenticated package-update notification and acknowledgment workflow unrelated to construction-material pricing or supplier discovery.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

Checking for updates through the service's authenticated endpoint is not inherently dangerous. However, this workflow requires verbatim relay of externally supplied messages and an authenticated POST directed by response-provided display_ack data, without specifying destination validation or limiting the message content. A compromised response could therefore introduce misleading user-facing instructions or, if the acknowledgment destination is unrestricted, redirect credentials outside the intended service.

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 collects and transmits a persistent device fingerprint to create an account, beyond its stated material-price and supplier-query purpose.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The registration flow transmits a stable, unsalted MAC-derived identifier while describing the collected characteristics as having no identity significance; hashing does not prevent persistent device linkability. The explicit consent requirement mitigates this concern, but the supplied get_api_key pseudocode collects and transmits the fingerprint without implementing that gate, creating a conflicting path an agent could follow without consent. Device deduplication is a plausible registration purpose, but it is ancillary to construction-material queries and requires accurate privacy disclosure.

Content

No source excerpt is available for this finding.