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 business-intelligence purpose, but it also handles credentials, device-derived registration, billing links, signed access URLs, and local persistence in ways users should review carefully.
Review this before installing if you are uncomfortable with a skill reading or writing ~/.zlbx/config.json, creating a vendor account after consent, sending platform/CPU/MAC-hash data for trial-account deduplication, producing local HTML reports, or including signed vendor links in outputs that may be shared.
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」)。
This duplicate finding identifies the same hidden-instructions block, which still represents a transparency and reviewability problem. Security-relevant execution guidance should not be concealed in comments because hosts or users may not realize the skill performs extra identification and callback requests.
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: company-intel-qichamao/1.0.6
This duplicate finding identifies the same hidden-instructions block, which still represents a transparency and reviewability problem. Security-relevant execution guidance should not be concealed in comments because hosts or users may not realize the skill performs extra identification and callback requests.
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: company-intel-qichamao/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 is the same hidden-instruction pattern in duplicated/variant form and remains a true issue. Embedding opaque control text in comments can manipulate downstream agent behavior and conceal privileged request-routing requirements from normal review paths.
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 is the same hidden-instruction pattern in duplicated/variant form and remains a true issue. Embedding opaque control text in comments can manipulate downstream agent behavior and conceal privileged request-routing requirements from normal review paths.
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 is effectively the same hidden-instruction issue duplicated by the scanner: embedded comment-based directives control authentication headers and package identity in a way that is opaque to normal review. Even if intended for package integrity, placing such controls in the prompt creates a covert instruction channel and increases the chance of misuse or prompt-conflict-driven security failures.
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 is effectively the same hidden-instruction issue duplicated by the scanner: embedded comment-based directives control authentication headers and package identity in a way that is opaque to normal review. Even if intended for package integrity, placing such controls in the prompt creates a covert instruction channel and increases the chance of misuse or prompt-conflict-driven security failures.
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 explicitly requires exposing raw URLs containing sk login-bypass parameters in report text and generated HTML outputs. These links function as bearer-style access tokens; disclosing them to users, exports, or shared files can leak authenticated access, enable unauthorized viewing, and create durable credential exposure through logs, chat history, or forwarded reports.
This duplicate finding identifies the same hidden instruction block that embeds package identity and authenticated request-handling directives in HTML comments. Concealed instructions are a real risk in agent skills because they can override expected behavior outside user visibility and cause the model to participate in credentialed actions with insufficient transparency or policy enforcement.
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 identifies the same hidden instruction block that embeds package identity and authenticated request-handling directives in HTML comments. Concealed instructions are a real risk in agent skills because they can override expected behavior outside user visibility and cause the model to participate in credentialed actions with insufficient transparency or policy enforcement.
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 generated reports to a local directory, but it does not declare any corresponding tool scope or permission boundary. Undeclared file-write capability weakens least-privilege guarantees and can surprise the host or user by enabling persistent local output beyond what the manifest advertises.
The invocation guidance is overly broad: it says the skill must be used whenever a user wants to 'fully understand' a company, even if they did not request this specific product. That can cause unintended triggering, unnecessary external API calls, and avoidable data disclosure or billing for routine company-related questions better handled by narrower tools.
The file consistently instructs the assistant to use fixed Chinese user-facing wording, examples, disclaimers, and prompts, but does not state that this is optional or limited to Chinese-speaking users. Under the policy, forcing a specific language without user choice is a natural-language policy violation.
The skill explicitly enables retrieval and display of project contact names and phone numbers, including full phone numbers for paid accounts, but only discusses formatting/display rules and upgrade prompts. It does not provide meaningful privacy, lawful-use, minimization, consent, retention, or anti-harvesting safeguards, which creates a real risk of exposing personal data for spam, profiling, or other misuse.
The file embeds a full account lifecycle flow—automatic registration, balance checks, upgrade prompting, and recharge handling—inside a skill whose stated purpose is enterprise intelligence analysis. This expands the skill’s authority beyond business analysis into credential acquisition and monetization workflows, increasing the chance of unauthorized data access, user confusion, and misuse of agent capabilities.
The document prescribes user-visible text entirely in Chinese, including a reusable privacy disclosure and later required output copy, but does not offer any user language selection or opt-in. This creates a natural-language locale policy issue because the skill effectively mandates one language regardless of the user's preferred language.
The skill instructs collection of device fingerprinting inputs, including a hashed MAC-derived identifier, for automatic account creation even though that is unrelated to the enterprise-intelligence task the user requested. Hardware-derived identifiers are sensitive telemetry, and collecting/transmitting them creates privacy and tracking risk disproportionate to the skill’s declared purpose.
The skill directs the agent to inspect local environment/config credentials and to persist newly obtained API keys into a user config file. Reading local secrets and writing auth material to disk materially broadens the skill’s privilege surface beyond analysis, creating risk of credential misuse, accidental overwrite, persistence of untrusted keys, and crossing trust boundaries on the host system.
These lines provide a mandatory example consent prompt for the agent to present to users, and the surrounding instructions imply this wording can be used directly. Because no alternative language behavior is provided, the skill enforces a specific language for a core user interaction without opt-in.
This instruction explicitly prepares and sends collected device features to an external service for auto-registration. Because the transmitted payload includes hardware-derived fingerprint data and account bootstrap metadata unrelated to the user’s requested analysis, the external transmission creates a real privacy and data-governance risk.
> ### ⚠️ 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 -d reference by itself is only an example of how to send the same registration payload and is not independently dangerous beyond the underlying external transmission already identified. As a separate finding it is mostly duplicative, but it still reinforces instructions for sending collected data off-device.
> ### ⚠️ 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