Back to skill

Security audit

建筑工程标讯洞察-筑龙标事

Security checks for vulnerabilities and agentic risk

Overview

The skill provides construction-tender research with disclosed account access, optional device-based signup requiring consent, and saved credentials.

Queries use service credits and send your search terms to the provider. Supply your own ZLBX_API_KEY to skip device-based signup. If you choose automatic signup, understand that hashing still produces a stable device identifier. Protect the saved key and generated login links. The scanner's partial-reference gaps were investigated; this review does not verify the provider's server-side privacy practices.

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 (9)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied code derives a hardware-based identifier rather than querying construction projects or recommending suppliers. Local device fingerprinting is not disclosed in the description, and the snippet establishes no connection to a necessary supporting function. It shows no 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
96% confidence
Finding

The supplied code performs local device identification rather than tender tracking, supplier recommendations, or project-deadline queries. The declared purpose does not disclose device fingerprinting, and the chunk provides no evidence that it is necessary to support the advertised functionality. No transmission or exfiltration of the identifier is shown.

Content

No source excerpt is available for this finding.

The credential setup workflow collects and transmits device fingerprints unrelated to construction-bid analysis.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L052–L056 instruct the agent to collect platform, architecture, and a MAC-address hash and submit those device characteristics to an account-registration endpoint. Construction-project tracking and potential-bidder recommendations justify authenticated data queries, but not hardware-derived device identification. The required user consent mitigates surprise but does not make this capability necessary for the manifest's stated purpose.

Content

No source excerpt is available for this finding.

The registration workflow collects and transmits a persistent hardware fingerprint, a capability outside the stated bid-insight purpose.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

The commands at L70–L98 obtain a network adapter's MAC address and hash it, and the registration request at L109–L125 transmits that hash with platform and architecture information. Hardware-based trial-account deduplication is not a direct requirement of infrastructure-project tracking or potential-bidder recommendations. The consent requirement mitigates undisclosed collection but does not make device fingerprinting part of the manifest's stated 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**

L097–L099 introduce an automatic package-update notification and acknowledgement workflow unrelated to construction-bid intelligence.

Low
Category
Excessive Agency
Confidence
88% confidence
Finding

The instructions require a balance request on the first use in every session to check package-update notices, followed by an authenticated POST acknowledging any notice displayed. This adds a package-maintenance capability that changes remote notification state, rather than supporting project tracking or potential-bidder discovery described in the manifest.

Content

No source excerpt is available for this finding.

L334–L336 require an unrelated package-update workflow, including an account query and a state-changing acknowledgement.

Low
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest describes construction-project intelligence and potential-bidder discovery, but these instructions require checking package-update notices through an authenticated account endpoint on first use each session. They also require posting a display acknowledgement that changes the server's pending-notice state; neither operation is necessary to search projects or recommend suppliers.

Content

No source excerpt is available for this finding.