Back to skill

Security audit

能源电力采招分析仪-电力招标网

Security checks for vulnerabilities and agentic risk

Overview

The procurement features are reasonable, but automatic update acknowledgments leave unclear limits on requests made with your API key.

Review the update workflow before installing: acknowledgments should use a fixed provider endpoint, and update messages should remain untrusted text. Supply an existing API key to skip device fingerprinting. Searches disclose your research targets to the provider, and paid queries consume credits. All bundled files were inspected; live server responses and server-side data handling 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 (13)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code performs host fingerprinting rather than procurement aggregation or award-concentration analysis. Reading and hashing a local MAC address is an undeclared capability with no evident supporting role in the stated purpose. The chunk does not show transmission of the resulting identifier.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied command performs local device identification rather than collecting or analyzing energy procurement data. This hardware-information access is not described or evidently necessary as a supporting implementation detail of the declared purpose. The snippet does not show transmission of the resulting hash.

Content

No source excerpt is available for this finding.

The skill collects and transmits device fingerprints for account registration, beyond its stated energy-procurement analysis purpose.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

L052–L056 direct the agent to collect platform, architecture, and a MAC-derived hash and submit them to an account-registration endpoint. Authenticating procurement queries is expected, but collecting device identifiers to provision an account is a separate capability not justified by the manifest's aggregation and award-concentration analysis purpose. The required consent reduces the privacy risk but does not eliminate the purpose mismatch.

Content

No source excerpt is available for this finding.

L097–L099 require an unsolicited upgrade-notification workflow and remote acknowledgment beyond energy procurement analysis.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

The manifest describes aggregating energy procurement projects and analyzing award concentration, but these instructions require an authenticated account request at the first use in every session to check upgrade notices. They also require relaying notices and submitting a POST acknowledgment that updates server-side notification state, even when the user requested only procurement analysis.

Content

No source excerpt is available for this finding.

Automatic company-search workflow sends user-supplied company names to an external service 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 use the matched company list in subsequent API queries without interrupting the user. Unlike a passive API catalog entry, this is an instruction to transmit user-provided research targets to a third-party service, but it provides no user-facing notice about that disclosure.

Content

No source excerpt is available for this finding.

The skill exposes personal contact lookup beyond its stated energy-procurement concentration analysis purpose (L220–L237).

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

The manifest describes aggregating energy-sector procurement projects and analyzing award concentration for equipment or engineering work. The documented get_company_contacts capability instead retrieves individuals’ names and telephone numbers, including complete numbers for paid accounts; these personal contact details are unnecessary for the stated analysis, even with the documented masking and bulk-export restrictions.

Content

No source excerpt is available for this finding.

Mandatory account and upgrade-notification workflow is unrelated to energy procurement analysis.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L393 requires an authenticated account-balance request at the first use of the skill in each session to check upgrade notices. L394–L395 additionally require forwarding those notices and submitting a POST display acknowledgment, introducing account-management and remote notification-state operations that are not needed to aggregate energy procurement projects or analyze award concentration.

Content

No source excerpt is available for this finding.

The skill defines paid corporate annual-report retrieval unrelated to its stated energy-procurement concentration analysis.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

The reference specifies an annual-report capability returning company phone numbers and email addresses, financial results, social-security information, and external guarantees. This general corporate due-diligence capability goes beyond aggregating energy procurement projects and analyzing award concentration, without a stated connection to that purpose. Successful requests also incur the charges described at L024.

Content

No source excerpt is available for this finding.

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

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The workflow requires verbatim display of server-controlled update messages and an authenticated acknowledgment based on response-provided instructions. Without validation and destination restrictions, this gives remote content an opportunity to introduce misleading upgrade instructions or steer authenticated actions outside the procurement-analysis task; the excerpt does not establish actual credential theft.

Content

No source excerpt is available for this finding.

The skill collects and transmits a hardware-derived device fingerprint for trial-account registration, beyond its stated procurement-analysis purpose.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The manifest describes aggregating energy-sector procurement projects and analyzing award concentration. The instructions instead include reading local network-interface MAC addresses, hashing them (L075–L099), and sending the hash with OS and architecture information to an account-registration endpoint (L109–L125). Although the workflow requires user consent, persistent device fingerprinting for trial eligibility is a separate capability not declared in the manifest or directly required for procurement analysis.

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**