Back to skill

Security audit

标前分析助手-投标前必查的一份报告

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to support bid analysis, but it deserves Review because it can create and store account credentials, collect a device-derived identifier, and place signed access links into shareable reports.

Install only if you are comfortable using this vendor's procurement API, allowing the skill to read or store a ZLBX API key, and potentially creating a trial account tied to a hashed device identifier. Treat generated sk and auto-login links as sensitive: avoid sharing reports or screenshots containing those URLs unless you intend recipients to have access.

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 (32)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

声明描述的是一个面向投标前调查与决策分析的核心分析型 skill;而提供的代码片段是一个展示层/导出层工具,职责是把已经生成好的分析结果 JSON 转成带样式的 HTML 报告。代码中没有网络请求、数据库访问、搜索、抓取、模型分析、规则校验或任何尽调计算逻辑。它处理的字段(如 buyer_analysis、competitors、pricing、risks 等)都假定已在输入 JSON 中准备好,脚本只是条件渲染这些内容。因此,代码的实际主要用途与声明的主要用途存在实质性偏差。虽然渲染报告可视为整体产品链路中的支持环节,但若用该代码代表该 skill 的行为,则明显不足以支撑所宣称的分析能力,属于描述与行为不匹配。

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill uses very broad mandatory activation language, requiring use for loosely defined bidding-related requests. Overbroad triggers can cause the agent to invoke this skill in situations the user did not clearly intend, leading to unnecessary external API calls, unintended data disclosure to the vendor API, and avoidable quota consumption.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 81)May include surrounding context.

md
完整模式下,对话报告输出后默认再用 `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 85)May include surrounding context.

md
完整模式下,对话报告输出后默认再用 `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 duplicate finding points to the same hidden-instruction block that governs request headers and post-display acknowledgment behavior. Concealed control logic increases supply-chain and privacy risk because it can silently alter external requests or enforce vendor-specific tracking semantics without clear user awareness.

Content

Scanner excerpt · SKILL.md (reported line 163)May include surrounding context.

  • 报告里某个竞争对手威胁最大 → 引导用 zlbx-company-intel 对其深度背调
  • 以上都不贴切 → 报告涉及的企业完整档案与更多商机详情,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: pre-bid-analysis-assistant/1.0.9

Hidden Instructions

High
Category
Prompt Injection
Confidence
93% confidence
Finding

This duplicate finding points to the same hidden-instruction block that governs request headers and post-display acknowledgment behavior. Concealed control logic increases supply-chain and privacy risk because it can silently alter external requests or enforce vendor-specific tracking semantics without clear user awareness.

Content

Scanner excerpt · SKILL.md (reported line 163)May include surrounding context.

  • 报告里某个竞争对手威胁最大 → 引导用 zlbx-company-intel 对其深度背调
  • 以上都不贴切 → 报告涉及的企业完整档案与更多商机详情,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: pre-bid-analysis-assistant/1.0.9

Hidden Instructions

High
Category
Prompt Injection
Confidence
95% confidence
Finding

This is the same underlying issue: hidden instructions in the skill content direct the agent to use specific authentication-related headers and workflow steps. In a security-sensitive agent environment, concealed prompt directives can manipulate request construction and induce unauthorized or unnecessary external interactions, especially when they involve credentials or account state checks.

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: bid-decision/1.0.5

Hidden Instructions

High
Category
Prompt Injection
Confidence
95% confidence
Finding

This is the same underlying issue: hidden instructions in the skill content direct the agent to use specific authentication-related headers and workflow steps. In a security-sensitive agent environment, concealed prompt directives can manipulate request construction and induce unauthorized or unnecessary external interactions, especially when they involve credentials or account state checks.

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: bid-decision/1.0.5

Hidden Instructions

High
Category
Prompt Injection
Confidence
94% confidence
Finding

This duplicate finding reflects the same hidden-instruction issue: covert prompt content dictates authentication/header behavior and package provenance. Such concealed directives are especially risky in a skill that also handles credentials and external requests, because they reduce transparency and auditability.

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: bid-decision/1.0.5

Hidden Instructions

High
Category
Prompt Injection
Confidence
94% confidence
Finding

This duplicate finding reflects the same hidden-instruction issue: covert prompt content dictates authentication/header behavior and package provenance. Such concealed directives are especially risky in a skill that also handles credentials and external requests, because they reduce transparency and auditability.

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: bid-decision/1.0.5

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The document embeds a full automatic account-registration and credential-handling workflow inside a skill whose stated purpose is pre-bid analysis. It instructs the agent to collect device-derived identifiers, contact external registration endpoints, persist API keys locally, and manage login/recharge flows, which materially exceeds the minimum permissions and behaviors expected for analytical assistance.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill directs collection of platform, CPU architecture, and a MAC-derived hash to build a device fingerprint and transmit it to a remote service. Even with hashing, this is persistent device tracking unrelated to the advertised procurement-analysis function, creating unnecessary privacy risk and enabling covert endpoint profiling.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
85% confidence
Finding

This is a duplicate hidden-instructions finding on the same comment block, and the concern is valid for the same reason: covert prompt content is being used to steer API request identity and behavior. Such hidden directives increase the risk of instruction smuggling and make security review harder because critical behavior is embedded in content rather than explicit configuration.

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: bid-decision/1.0.5

Hidden Instructions

High
Category
Prompt Injection
Confidence
85% confidence
Finding

This is a duplicate hidden-instructions finding on the same comment block, and the concern is valid for the same reason: covert prompt content is being used to steer API request identity and behavior. Such hidden directives increase the risk of instruction smuggling and make security review harder because critical behavior is embedded in content rather than explicit configuration.

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: bid-decision/1.0.5

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The instructions mandate sharing complete URLs with sk login-signature parameters and even forbid removing them, while giving no warning that these links are sensitive credentials. This creates a direct path for token leakage to end users, logs, screenshots, exported HTML reports, and any third party with access to the conversation.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
94% confidence
Finding

This duplicated hidden-instruction finding points to the same issue: the skill content instructs the agent to preserve and transmit specific identity headers and reuse configured credentials. Even if the destination appears legitimate, letting skill-authored content control authenticated request metadata is dangerous because a compromised or modified skill could cause unauthorized disclosure, request spoofing, or policy bypass.

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: bid-decision/1.0.5

Hidden Instructions

High
Category
Prompt Injection
Confidence
94% confidence
Finding

This duplicated hidden-instruction finding points to the same issue: the skill content instructs the agent to preserve and transmit specific identity headers and reuse configured credentials. Even if the destination appears legitimate, letting skill-authored content control authenticated request metadata is dangerous because a compromised or modified skill could cause unauthorized disclosure, request spoofing, or policy bypass.

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: bid-decision/1.0.5

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 71)May include surrounding context.

python
THREAT_COLOR = {"高": "#c0392b", "中": "#b9770e", "低": "#0d9463"}

# 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
88% confidence
Finding

The skill explicitly performs local file writes by generating and saving HTML reports, but it does not declare any scoped permissions or allowed-tools boundary. That creates a capability/permission mismatch: the runtime or reviewer cannot clearly constrain or audit where and when file output is allowed, increasing the risk of unintended local writes or abuse if the skill is modified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
78% confidence
Finding

The file consistently instructs behavior and example outputs entirely in Chinese for all user interactions, but does not offer the user a language or locale choice. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The document explicitly states that the privacy notice should be relayed to the user as-is, and the entire set of required user-facing prompts and messages is written in Chinese. There is no indication that the agent should adapt to the user's preferred language or ask for consent to use Chinese, which is a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file instructs the agent to inspect environment variables and local configuration files for API credentials, then use and later persist keys. Reading and writing secrets from user-local storage is a privileged action unrelated to bid analysis and increases the risk of credential misuse, cross-skill leakage, and unintended persistence.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

This instruction describes sending collected device features and registration data to an external endpoint. In context, the transmission is not necessary for the skill’s stated analytical purpose and forms part of an unauthorized data-exfiltration and account-enrollment workflow.

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
91% confidence
Finding

The curl-based guidance is another explicit path for transmitting locally collected data to a remote registration service. Although framed as implementation advice, it still operationalizes external data transfer unrelated to the advertised functionality.

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
95% confidence
Finding

This duplicate finding concerns the same session-persistence behavior: local directory creation, config mutation, and durable API-key storage. In the context of an analysis skill, this persistence is not justified and materially expands the blast radius of any compromise or misbehavior.

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**