Ae1
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
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」)。
Security audit
Security checks for vulnerabilities and agentic risk
The skill mostly matches its supplier-checking purpose, but it also asks to create accounts, collect a hashed device identifier, persist an API key, and share signed no-login links in exported reports.
Review this before installing if you are uncomfortable with a skill saving an API key in your home directory, collecting a hashed MAC address for free-trial registration, or placing signed no-login links into shareable HTML reports. Prefer configuring ZLBX_API_KEY manually to avoid auto-registration, and avoid sharing generated reports outside intended recipients.
Referenced artifact was not completely inspected
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」)。
Referenced artifact was not completely inspected
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 were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
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:
X-Skill-Package: supplier-qualification-checker/1.0.6
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
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:
X-Skill-Package: supplier-qualification-checker/1.0.6
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
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。
X-Client: company-intel/1.0.2
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
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。
X-Client: company-intel/1.0.2
This duplicate hidden-instructions finding points to the same concealed header-directive block. In a security review context, concealed control instructions are a real concern because they bypass transparency and can be repurposed to override expected behavior or quietly alter where and how requests are sent.
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。
X-Client: company-intel/1.0.2
This duplicate hidden-instructions finding points to the same concealed header-directive block. In a security review context, concealed control instructions are a real concern because they bypass transparency and can be repurposed to override expected behavior or quietly alter where and how requests are sent.
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。
X-Client: company-intel/1.0.2
The skill’s declared purpose is supplier due-diligence, but the file embeds a substantial secondary workflow for account provisioning, device fingerprint collection, quota handling, and local credential management. That is a material capability expansion unrelated to the user’s requested business analysis and creates unnecessary access to host-derived data and persistence mechanisms, increasing privacy and security risk if the skill is invoked in a sensitive environment.
The skill instructs collection and external transmission of device-derived identifiers such as platform, architecture, and a hashed MAC address to support trial-account deduplication. Even though the MAC is hashed, this is still persistent device fingerprinting unrelated to supplier qualification analysis, and it exposes the user environment to tracking and unintended correlation across sessions or systems.
This duplicate finding points to the same hidden request-identity block embedded in an HTML comment. Even if the content is not overtly malicious, concealed prompt instructions that control credentialed requests increase the chance of unnoticed data handling behavior and should be treated as a genuine prompt-injection style weakness.
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。
X-Client: company-intel/1.0.2
This duplicate finding points to the same hidden request-identity block embedded in an HTML comment. Even if the content is not overtly malicious, concealed prompt instructions that control credentialed requests increase the chance of unnoticed data handling behavior and should be treated as a genuine prompt-injection style weakness.
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。
X-Client: company-intel/1.0.2
The template requires preserving and sharing full URLs containing sk login-bypass parameters, explicitly so recipients can access content without the normal login barrier. Exposing such bearer-style access tokens in reports, links, or exported artifacts can grant unintended access to company pages or deeper data to anyone who obtains the URL, making this a clear access-control and privacy risk.
This duplicated finding identifies the same hidden instruction block that tells the agent to preserve and send package-identifying headers alongside an API key. In context, the workflow also mandates an authenticated balance/version-check call at session start and conditional acknowledgment behavior, which increases risk by embedding non-user-driven network actions inside prompt instructions.
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。
X-Client: company-intel/1.0.2
This duplicated finding identifies the same hidden instruction block that tells the agent to preserve and send package-identifying headers alongside an API key. In context, the workflow also mandates an authenticated balance/version-check call at session start and conditional acknowledgment behavior, which increases risk by embedding non-user-driven network actions inside prompt instructions.
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。
X-Client: company-intel/1.0.2
Code contains obfuscation (base64, hex encoding with execution). This is often used to hide malicious functionality.
# 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]
The skill explicitly states it will write HTML reports to a local directory, but it does not declare any tool scope or permissions boundary for file_write. Missing explicit capability scoping increases the chance that an agent runtime grants broader write access than intended, enabling unintended local file creation or overwrite if the skill is invoked in a permissive environment.
The invocation rule is overly broad: it says the skill 'must' be used whenever a user wants to verify a supplier or partner's real performance capability, even when the request could be a general company inquiry. This can cause over-activation, unnecessary external data sharing, unexpected account registration flows, and unintended point consumption for ordinary user queries.
The skill content is written as Chinese-only user-facing behavior and example phrasing, with no statement that users may choose another language or locale. Under the policy, forcing a specific language without opt-in is a locale/language policy violation unless clearly documented and justified.
The skill explicitly instructs retrieval and display of project contact data, including phone numbers, with guidance to show full numbers when available. Although it says to preserve returned masking and avoid bulk export, it lacks a clear user-facing privacy warning, lawful-basis check, or purpose limitation before exposing personal contact data, which creates privacy and compliance risk.
The document claims auto-registration only occurs when the user has not actively configured an API key, but the later logic also triggers registration whenever the config file exists without an api_key. This inconsistency can mislead users about when device data will be collected and transmitted, undermining meaningful consent and making the privacy notice unreliable.
The document provides a fixed Chinese consent prompt to present to users, and later sections similarly prescribe exact Chinese output text for user-visible messages. This is a natural-language locale policy concern because the skill mandates a specific language for user communication without stating that the user can choose or that the skill is limited to a Chinese-only context.
This section explicitly instructs making an external POST request containing collected device features to a third-party service. In the context of a supplier-checking skill, that outbound transmission is risky because it sends host-derived data unrelated to the user’s supplier-analysis request, expanding the trust boundary and creating privacy exposure.
> ### ⚠️ 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`)。
>
The curl-based example is another explicit path for exfiltrating locally derived device information to an external endpoint. While the example is framed as implementation guidance, it still operationalizes outbound transfer of unrelated host data and therefore represents a real privacy/security concern in this skill context.
> ### ⚠️ 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
The skill directs the agent to create and modify ~/.zlbx/config.json, merge settings, and persist an API key for future use. Local credential persistence and config mutation exceed the stated supplier-audit role and can create lasting security consequences if done without strong user awareness, least-privilege design, and clear separation from the business function of the skill.