Back to skill

Security audit

企业情报-招投标视角的企业背调

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent company-intelligence skill, but users should understand that optional trial signup sends a hashed device identifier to the vendor and stores an API key locally.

Install only if you are comfortable using the vendor's API, storing an API key in a local config file, and, if you do not provide your own key, approving trial signup that sends platform, CPU architecture, and a hashed MAC identifier for quota deduplication. Use contact lookup only for legitimate business purposes.

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 (33)

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
90% confidence
Finding

This duplicate finding points to the same hidden instruction block. Even if the specific headers are legitimate package identity fields, concealing mandatory request behavior and confirmation logic in comments is dangerous because agents may obey non-obvious instructions that bypass user awareness and complicate security review.

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: zhiliao-company-intel/1.0.6

Hidden Instructions

High
Category
Prompt Injection
Confidence
90% confidence
Finding

This duplicate finding points to the same hidden instruction block. Even if the specific headers are legitimate package identity fields, concealing mandatory request behavior and confirmation logic in comments is dangerous because agents may obey non-obvious instructions that bypass user awareness and complicate security review.

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: zhiliao-company-intel/1.0.6

Hidden Instructions

High
Category
Prompt Injection
Confidence
97% confidence
Finding

This is a duplicate hidden-instructions finding on the same comment-embedded identity block, and the underlying issue is real: the skill embeds non-user-visible directives about credentialed request headers and package provenance. In a security review, such hidden control text is risky because it can manipulate agent behavior without transparent review by the end user.

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
97% confidence
Finding

This is a duplicate hidden-instructions finding on the same comment-embedded identity block, and the underlying issue is real: the skill embeds non-user-visible directives about credentialed request headers and package provenance. In a security review, such hidden control text is risky because it can manipulate agent behavior without transparent review by the end user.

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
95% confidence
Finding

This duplicate finding reflects the same hidden-instruction issue: invisible comment-based directives alter request identity and handling. Because such instructions are embedded in prose rather than enforced by trusted code or manifest policy, they create a channel for covert behavior and policy bypass attempts.

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
95% confidence
Finding

This duplicate finding reflects the same hidden-instruction issue: invisible comment-based directives alter request identity and handling. Because such instructions are embedded in prose rather than enforced by trusted code or manifest policy, they create a channel for covert behavior and policy bypass attempts.

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 directs collection of host-level device fingerprints, including OS, architecture, and a hashed MAC-derived identifier, then sends them to a remote service for trial-account registration. Even with hashing, this is still persistent device tracking unrelated to the core company-intelligence purpose, and it creates privacy risk plus a precedent for local host reconnaissance inside an otherwise data-query skill.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
98% confidence
Finding

This duplicate hidden-instructions finding is also valid: the skill embeds concealed directives that force use of specific API identity headers and source/version values. Even if the values are legitimate, placing them in hidden prompt content is risky because the model may treat them as authoritative operational commands, enabling stealthy manipulation of outbound authenticated requests beyond the visible user task.

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
98% confidence
Finding

This duplicate hidden-instructions finding is also valid: the skill embeds concealed directives that force use of specific API identity headers and source/version values. Even if the values are legitimate, placing them in hidden prompt content is risky because the model may treat them as authoritative operational commands, enabling stealthy manipulation of outbound authenticated requests beyond the visible user task.

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 duplicate hidden-instructions finding is also valid: the skill embeds non-user-visible directives telling the agent exactly which identity headers to send and to preserve an already configured API key. Such covert control is risky because skill content is adversarial input; if normalized, it can be used to manipulate authentication context, leak package provenance, or override safer platform defaults without transparent disclosure to the user.

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 duplicate hidden-instructions finding is also valid: the skill embeds non-user-visible directives telling the agent exactly which identity headers to send and to preserve an already configured API key. Such covert control is risky because skill content is adversarial input; if normalized, it can be used to manipulate authentication context, leak package provenance, or override safer platform defaults without transparent disclosure to the user.

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 = "iVBORw0KGgoAAAANSUhEUgAAASsAAABgCAYAAABFTFSvAAAAAXNSR0IArs4c6QAAAERlWElmTU0AKgAAAAgAAYdpAAQAAAABAAAAGgAAAAAAA6ABAAMAAAABAAEAAKACAAQAAAABAAABK6ADAAQAAAABAAAAYAAAAACurZhBAAA2V0lEQVR4Ae2dB5xU1fXHzzZ6752ldxHpHTtqTKJJjEmMGkvKP/aIRk0UayzYxZJouokaYyzRiKCAoIB0kN47LB0WFnZh9//9vZmHb4eZnZndmWVW3vl8fjOv3fvuO+++3z3nvHvvM/PF14CvAV8DFUADafbM2OusKOMCKywoqgDlDV/ENIpemJ5vGYW5VmjbrChtlaXZYks7usRuumlb+ET+Vl8DvgYqkgYyzdJ6WqWskXa0IhU7UlnTzLgiK4K8jh7litK22FPPT4O03rb8zAk26hc5kVL6230N+BpIbQ1kWpEdsYJ8Hu4jqV3SeEuXlpZBkhaWmf49s/RvW9aRFfbs2D9ZweF/2i23bIo3O/94XwO+Bk6sBtJP7OmTeHZZV0JBgdmRI1mWnt7V0jIesozK7+H6/shGj66UxLP7Wfsa8DWQYA18fcnKq6iAWxggrbS0XpDWc1a34dP2zDMtvIf5y74GfA2krgZODrJy9e+SVuHROpae8ROC8i/bsy/1cHf7/74GfA2krgZOLrJy70NhoWJ0lXENz8Tces6eHHuqu8v/9zXgayA1NXBykpXuhayswkJeMKQPtHR7greGnVLzFvml8jXga0AaOHnJSlfvENbRLEtLG2BphaPtiSfqabMvvgZ8DaSeBk5ustL9EGEVFVW1jIwzLC3rlyzTWcsXXwO+BlJNAz5Z6Y4ohnXkaCPLSP++PfnCkFS7SX55fA34GjjZ3cDjakBatqUduQZ3sOpxu/wNvgZ8DZxQDfiWlav+gDtY3TKz+plVHupu9v99DfgaSA0N+GTlvQ8Bwmpm6YUX2htvaLiOL74GfA2kiAaST1bEr5Miycq3sLAm5e1jG7e3SUq5/Ux9DfgaKJUGkkNWRxgUfUSTHjAJQjqnKGRZ6wpklyQByybyEZpIQeC4jHQy1/HKV9sSIcovLS0NV7ARhcUd9MXXgK+BVNGAJlRJnEBGVbOy7Lz2be1bbdtax3p1LYtnf1tenk3euMneWrHSVu7cbTCNSAGyUU9yCILV9IxMq5KVaUfZdPioBh9DQDomTXxaZNUyM613s2Z2UYe2dlrDhlaD8+w5nG8zt22zd1atthlbmcZKZChyLKsUFtWxtKJTyOYfZc3KT+9rwNdAYjSQZk8//wKk8POyThGTAfF0b9DQHh022M5p3eq40slr27R/vz06a649N28uYSGzVnXr2vltWtvpLVpY29q1rHqlLDsC4ew8dMjm5+yw91avsbk5260mxHRr3952eZdOVp3lUNHxf1m01B6bNdu25uZChmUINzkEmXYIa+1dGPNKppPJCz2fv+5rwNdA+WsgIZZVBpbPKQ0b2avnn2tdsKaO4E6t3L3bpm/ZagcKjjgW1sAmTaxFzZr20JCB1q1BPQwXw0pqZw2rhu8lMBQr6mendLecgwctl2leOkFsynfdvn32xdYcZ3vzmjWsb+NG1rxGDefYJtWr2k2Tpth20pTJwioqghHTGllWVl1uiU9W5V8v/TP6GjhOAwkhq/pVqth9gwY4RLU3P98enTnHnsd62pN32IkrZeLi9Wve1P5w1unWtX59+1mP7scVJNyGLFw6EZGkEKKat32H3Txxkk1du94ho0xcQ+V3W7/e9oNOHe2c7NZ2Q699ds/n05nduJTCeZAM4m01sK4UbPfF14CvgRTQQJnJqjJhpcHNm9t5uHP78wvsqTlz7SHIwnHFIBMFwasSixrOMY2qVS92yUcKi2zFnt02f/tOW757j23PO2hZxL
...[truncated 27 chars]

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill explicitly writes HTML reports to a local directory, but it does not declare any tool scope or permissions boundary for file-write capability. This weakens least-privilege controls and can allow an agent runtime to grant broader filesystem access than users expect, increasing the blast radius if the skill is misused or later modified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest description is written as a mandatory instruction that this skill must be used for certain requests, and all user-facing examples and operational guidance are in Chinese with no opt-in or alternative locale path. Under the policy, forcing a specific language or locale without user choice is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The file explicitly instructs the agent to attach an API key and package identity headers to outbound requests, but it does not include any user-facing warning or consent language about credential use, third-party transmission, or trust boundaries. In a skill package, this increases the risk of silent exfiltration of user-linked credentials or metadata to an external service, especially because the content must be treated as untrusted prompt material.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The contact lookup section directs retrieval and display of phone numbers, including full numbers for paid accounts, without a prominent warning about personal data sensitivity, lawful basis, or usage restrictions. That creates a privacy and compliance risk because the skill normalizes access to personal contact data and instructs disclosure in chat output.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill’s documented behavior expands well beyond enterprise-intelligence reporting into account provisioning, credential handling, recharge-link generation, and session/account lifecycle management. That functional overreach increases the attack surface and grants the skill authority to persist secrets and steer users through external account flows that are not necessary for the declared business-intel task.

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 is written as prescriptive user-facing guidance in Chinese, including text to relay verbatim to users, but it does not provide any option to match the user's preferred language. This can violate language/locale policy because the skill effectively mandates a specific language for disclosures and prompts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

These lines instruct the agent to obtain consent using a fixed Chinese sentence example, with no indication that the wording should be localized to the user's language. Because this is user-facing operational text, it represents a natural-language locale constraint without opt-in.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
92% confidence
Finding

This instruction governs POSTing collected device features to an external endpoint for automatic registration. The danger is not JSON serialization itself, but the sanctioned exfiltration of locally collected host identifiers to a third-party service from within a skill whose declared purpose is company-intelligence analysis.

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
90% confidence
Finding

The curl-based guidance reinforces direct external transmission of device-derived data and account-registration payloads. In context, this enables the skill to push local system-derived identifiers to a remote service, which is unnecessary for normal company-intelligence lookups and broadens privacy and data-handling risk.

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
93% confidence
Finding

The skill instructs the agent to write API credentials to ~/.zlbx/config.json and reuse them across sessions. Persisting secrets locally from within a content-retrieval skill exceeds expected scope, risks credential exposure on shared systems, and normalizes secret management by untrusted skill content.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
95% confidence
Finding

This duplicate finding identifies the same persistence behavior: creating a durable local config entry and reusing it automatically. In context, that turns a one-time skill invocation into a long-lived credential-management agent, which is materially more dangerous than transient API use.

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**