Back to skill

Security audit

招标中标查询&数据分析-乙方投标助手

Security checks for vulnerabilities and agentic risk

Overview

The skill supports tender research with disclosed API access, optional device-based registration requiring consent, and ordinary credential storage.

Use an existing API key to skip device registration, or review the consent prompt before sharing a stable MAC-derived identifier. Queries can consume paid credits. Protect the saved key and generated login links, and expect automatic update-notice checks and display receipts. All bundled files were inspected; the provider's server-side privacy 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
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (8)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code collects and hashes a local hardware identifier, which is unrelated to the declared tender-analysis functions and is not explained as a supporting implementation requirement. The supplied chunk does not transmit the identifier or perform registration, but its local hardware-identification behavior is undeclared.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied code performs device fingerprinting rather than tender-data retrieval or analysis. Local hardware identification is not disclosed in the declared purpose, and the chunk shows no connection to the bidding workflow that would establish it as a supporting implementation detail. No transmission, persistence, or use of the resulting hash is shown.

Content

No source excerpt is available for this finding.

The skill adds a mandatory package-update notification and acknowledgment workflow unrelated to bidding-data analysis.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

These instructions require an authenticated account request at the start of each session to check package upgrades, relay the returned notice, and submit a POST acknowledgment recording its display. Unlike balance checks needed to assess available query credits, this package-maintenance workflow introduces user-facing messaging and remote state changes outside the manifest’s bidding-research and decision-support purpose.

Content

No source excerpt is available for this finding.

Mandatory account-based upgrade checks and display acknowledgments extend beyond bid-data analysis.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

L393 requires an authenticated account request at the first use of every session specifically to check upgrade notices, and L395 requires a POST acknowledgment after displaying a notice. This introduces an automatic package-maintenance and remote acknowledgment workflow that is not needed for the manifest's tender searches, supplier analysis, or pricing support.

Content

No source excerpt is available for this finding.

Automatic registration collects and transmits a persistent hardware fingerprint beyond the bidding-analysis purpose.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The registration request sends platform, architecture, and a MAC-address hash to an external account-registration endpoint (L109–L124), using hardware collection commands specified at L65–L99. The stated purpose is procurement-data analysis, whereas this fingerprinting serves trial-account deduplication (L16–L17), not bidding analysis or ordinary API authentication. The explicit consent requirement mitigates the privacy concern but does not make hardware fingerprinting a declared capability of the skill.

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**