Back to skill

Security audit

启信宝企业背调-招投标视角查企业

Security checks for vulnerabilities and agentic risk

Overview

The skill largely matches its company-background-check purpose, but it also manages credentials, fingerprints the device for auto-registration, saves reports locally by default, and hides some service-call instructions in comments.

Review this skill before installing if you are uncomfortable with a third-party service receiving company queries, account registration metadata, and a hashed device identifier. Prefer setting your own ZLBX_API_KEY to avoid auto-registration, expect reports to be saved locally by default, and treat contact-phone lookups as personal-data handling.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
Findings (30)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 89)May include surrounding context.

md
默认再用 `scripts/render_report.py` 生成一份可分享的 HTML 版报告并告知保存路径(详见 report-template.md「HTML 报告导出」)。

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 93)May include surrounding context.

md
默认再用 `scripts/render_report.py` 生成一份可分享的 HTML 版报告并告知保存路径(详见 report-template.md「HTML 报告导出」)。

Hidden Instructions

High
Category
Prompt Injection
Confidence
93% confidence
Finding

This duplicated finding points to the same hidden-instructions block embedded in an HTML comment. Even if the content appears operationally benign, concealing behavior-affecting directives in comments undermines trust and can hide outbound request requirements or additional server interactions from standard review paths.

Content

Scanner excerpt · SKILL.md (reported line 185)May include surrounding context.

  • 用户想持续跟踪该公司 → 介绍竞对动态周报(见上方「竞对动态周报」)
  • 以上都不贴切 → 询问是否要对某个竞对做双公司对比,或引导访问知了商机大师 https://agent.zhiliaobiaoxun.com

安装包身份与升级检查:所有发往知了标讯 API 的请求(包括账户查询)必须额外携带以下两个请求头:

text
X-Skill-Package: company-background-check-qixinbao/1.0.6

Hidden Instructions

High
Category
Prompt Injection
Confidence
93% confidence
Finding

This duplicated finding points to the same hidden-instructions block embedded in an HTML comment. Even if the content appears operationally benign, concealing behavior-affecting directives in comments undermines trust and can hide outbound request requirements or additional server interactions from standard review paths.

Content

Scanner excerpt · SKILL.md (reported line 185)May include surrounding context.

  • 用户想持续跟踪该公司 → 介绍竞对动态周报(见上方「竞对动态周报」)
  • 以上都不贴切 → 询问是否要对某个竞对做双公司对比,或引导访问知了商机大师 https://agent.zhiliaobiaoxun.com

安装包身份与升级检查:所有发往知了标讯 API 的请求(包括账户查询)必须额外携带以下两个请求头:

text
X-Skill-Package: company-background-check-qixinbao/1.0.6

Hidden Instructions

High
Category
Prompt Injection
Confidence
84% confidence
Finding

This duplicated finding points to the same hidden-instruction block that tells the agent to use configured API credentials and attach package/source headers. Even if operationally legitimate, encoding these directives in hidden content creates an instruction-smuggling surface and can cause the model to prioritize non-user-visible instructions over safer control-plane mechanisms.

Content

Scanner excerpt · references/api-quick.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Hidden Instructions

High
Category
Prompt Injection
Confidence
84% confidence
Finding

This duplicated finding points to the same hidden-instruction block that tells the agent to use configured API credentials and attach package/source headers. Even if operationally legitimate, encoding these directives in hidden content creates an instruction-smuggling surface and can cause the model to prioritize non-user-visible instructions over safer control-plane mechanisms.

Content

Scanner excerpt · references/api-quick.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Hidden Instructions

High
Category
Prompt Injection
Confidence
90% confidence
Finding

This duplicate hidden-instruction finding is still valid because the same concealed block directs the agent to attach specific identity headers and not reuse other package sources. Such covert steering is inappropriate in a user-facing skill and can be abused to override normal request-routing or provenance controls.

Content

Scanner excerpt · references/auto-register.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Hidden Instructions

High
Category
Prompt Injection
Confidence
90% confidence
Finding

This duplicate hidden-instruction finding is still valid because the same concealed block directs the agent to attach specific identity headers and not reuse other package sources. Such covert steering is inappropriate in a user-facing skill and can be abused to override normal request-routing or provenance controls.

Content

Scanner excerpt · references/auto-register.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill instructs the agent to collect local device characteristics and run OS-specific shell/PowerShell commands to derive a MAC-based device fingerprint, then use it for remote account registration. For a company-intelligence skill, this is unnecessary host introspection and externalizes sensitive device-derived identifiers, creating privacy and fingerprinting risk even if the raw MAC is hashed.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
97% confidence
Finding

This duplicated finding points to the same hidden instruction block that tells the agent to reuse configured credentials and send package/source headers. In this skill context, that is especially sensitive because the template also instructs account-balance and acknowledgment calls, increasing the chance that an LLM follows untrusted embedded operational directions and performs authenticated requests it should not autonomously derive from prompt content.

Content

Scanner excerpt · references/report-template.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Hidden Instructions

High
Category
Prompt Injection
Confidence
97% confidence
Finding

This duplicated finding points to the same hidden instruction block that tells the agent to reuse configured credentials and send package/source headers. In this skill context, that is especially sensitive because the template also instructs account-balance and acknowledgment calls, increasing the chance that an LLM follows untrusted embedded operational directions and performs authenticated requests it should not autonomously derive from prompt content.

Content

Scanner excerpt · references/report-template.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Hidden Instructions

High
Category
Prompt Injection
Confidence
95% confidence
Finding

This duplicated hidden-instructions finding is still valid because the same HTML-comment block embeds imperative transport and identity directives at the top of the workflow. In this skill context, that is more dangerous because the package is specifically instructing credentialed calls to an external API, so concealed instructions directly affect how secrets and request identity are used.

Content

Scanner excerpt · references/workflow.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Hidden Instructions

High
Category
Prompt Injection
Confidence
95% confidence
Finding

This duplicated hidden-instructions finding is still valid because the same HTML-comment block embeds imperative transport and identity directives at the top of the workflow. In this skill context, that is more dangerous because the package is specifically instructing credentialed calls to an external API, so concealed instructions directly affect how secrets and request identity are used.

Content

Scanner excerpt · references/workflow.md (reported line 1)May include surrounding context.

请求身份:使用已配置的 API Key,并保留本包的 X-Client,携带下列完整身份头;不得沿用其它包的版本或来源。

text
X-Client: company-intel/1.0.2

Obfuscated Code

High
Category
Supply Chain
Confidence
50% confidence
Finding

Code contains obfuscation (base64, hex encoding with execution). This is often used to hide malicious functionality.

Content

Scanner excerpt · scripts/render_report.py (reported line 93)May include surrounding context.

python
# 知了标讯白色 logo(299x96 PNG base64 内嵌,保证报告离线/打印/转发时不裂图)
_LOGO_B64 = "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
...[truncated 27 chars]

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
85% confidence
Finding

The skill explicitly writes reports to local storage (~/zlbx-company-intel-files/) and instructs disclosure of the absolute path, but it does not declare a corresponding tool scope or permissions boundary. This creates a capability mismatch: users and hosts may not realize disk writes occur by default, increasing the risk of unintended local data persistence and weak auditability.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The activation language is very broad and says the skill must be used whenever a user wants to understand a company's background or strength, even if they did not ask for this workflow. That can cause over-triggering of a networked, billable, data-exporting skill on ordinary company questions, leading to unnecessary external disclosure, charges, and unintended workflow execution.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill instructs automatic generation of an HTML report and disclosure of its absolute local filesystem path, but the default disk write is not surfaced as a strong user-facing consent step. This can surprise users, leak sensitive local path information into the conversation, and persist potentially sensitive company analysis on disk without explicit opt-in.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill explicitly instructs the agent to retrieve and display project contact phone numbers, including full numbers for paid accounts, with only operational constraints about masking and export volume. In a company-background-check context, this enables disclosure of personal contact data without any stated legal basis, purpose limitation, consent check, or user-facing privacy warning, creating a meaningful privacy and compliance risk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill adds a full account lifecycle flow—auto-registration, credential issuance, persistence, and recharge handling—that is materially outside the declared company background-check purpose. This expands the skill’s authority into identity, billing, and credential management, increasing attack surface and creating opportunities for unauthorized account creation or misuse under the guise of a research tool.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
89% confidence
Finding

This section explicitly directs the agent to transmit collected device features and registration metadata to an external service. Because the transmitted data includes device-derived identifiers gathered from the local host for a purpose outside the skill’s stated function, the external transmission is privacy-sensitive and expands the risk of tracking or account abuse.

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
84% confidence
Finding

The curl-based example reinforces an external data exfiltration path for locally collected device features. While the serialization guidance itself is not the core issue, it operationalizes outbound transmission of unnecessary host identifiers from a company-research skill to a third-party endpoint.

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

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

Writing API keys into a persistent user config file and immediately reusing them in-session gives the skill credential-management powers unrelated to business due diligence. Persistent storage of secrets by a content skill increases the blast radius of compromise, can override user expectations, and may expose credentials to other local processes or future unintended reuse.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
95% confidence
Finding

This duplicate session-persistence finding is also valid because the same instructions direct the agent to persist credentials and make them immediately active in the current session. That combination creates durable access beyond the user’s immediate request and is not justified by the skill’s narrow research purpose.

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**

Session Persistence

Medium
Category
Rogue Agent
Confidence
95% confidence
Finding

This duplicate session-persistence finding is also valid because the same instructions direct the agent to persist credentials and make them immediately active in the current session. That combination creates durable access beyond the user’s immediate request and is not justified by the skill’s narrow research purpose.

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**

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The balance-exhaustion flow generates device SID auto-login links and guides users through recharge and phone-binding flows, which are unrelated to the stated research function. This introduces authentication-token issuance and account-recovery behaviors into a non-auth skill, raising phishing, session hijack, and abuse risks if mishandled.

Content

No source excerpt is available for this finding.