Back to skill

Security audit

travel-expense-faq

Security checks for vulnerabilities and agentic risk

Overview

This is a Chinese internal travel-expense FAQ skill with no code execution or persistence, but users should verify the reimbursement login uses HTTPS and handle bank and invoice data carefully.

Install only for users who need this specific Chinese corporate travel-expense guidance. Before using login instructions, confirm the official reimbursement portal through a trusted company source and use HTTPS; do not enter credentials, bank account numbers, or invoice data into an unverified site or expose them beyond the required finance workflow.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
references/费用报销PC端操作手册.md:1
Finding

Financial System Login Link Uses Plaintext HTTP

Content
View full analysis

Vulnerability Details

File Location: references/费用报销PC端操作手册.md, line 1
Vulnerability Type: Plaintext HTTP link for a financial system login workflow
Risk Level: Medium

Vulnerable Code

markdown
登录网址[www.jdy.com](http://www.jdy.com)

The text identifies www.jdy.com as the login address, while the actual hyperlink uses unencrypted HTTP.

Technical Analysis

The operating manual directs users to begin a financial-system login workflow through http://www.jdy.com. An initial HTTP request is neither encrypted nor authenticated. Even if the legitimate server normally redirects users to HTTPS, that redirect is received only after the vulnerable plaintext request has traversed the network.

An attacker capable of intercepting or modifying network traffic—such as an operator of a malicious Wi-Fi access point, a compromised gateway, or another on-path adversary—could replace the HTTP response before any legitimate HTTPS redirect occurs. The attacker could redirect the user to a visually similar phishing site or inject misleading login instructions.

This is relevant because the referenced system handles reimbursement records, invoices, and financial account information. The audit did not verify that the destination itself is malicious, and the project does not programmatically submit credentials or perform network requests. The issue is specifically the insecure protocol used by the documented login link.

Attack Path

  1. An employee opens the reimbursement manual and follows the documented link.
  2. The browser sends an unauthenticated plaintext HTTP request for www.jdy.com.
  3. An on-path attacker intercepts the request before the legitimate server responds.
  4. The attacker returns a redirect to a lookalike domain or serves a spoofed login page.
  5. The employee enters credentials or other sensitive information into the fraudulent page.
  6. The attacker captures that information and may attempt to access the legitimate reimbursemen ...[truncated 933 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace the plaintext URL with the service's verified canonical HTTPS login URL:

    markdown
    Login URL: [https://www.jdy.com](https://www.jdy.com)
    
  2. Confirm the canonical login address through the vendor's trusted documentation before publishing the replacement.

  3. Add guidance requiring users to verify the exact HTTPS hostname and valid certificate before entering credentials.

  4. Warn users not to proceed if the browser reports a certificate error, uses an unexpected domain, or falls back to HTTP.

  5. Where available, direct users through a trusted corporate portal or managed bookmark rather than a manually entered public URL.

  6. Enable multifactor authentication and phishing-resistant authentication controls for reimbursement accounts where supported.

  7. Search future documentation releases for http:// login or payment links and reject them during review unless there is a documented, non-sensitive justification.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (14)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The trigger list is broad enough to match generic reimbursement-related queries, which can cause the wrong skill to activate and provide travel-only guidance for non-travel expense cases. In an enterprise workflow, this can misroute user requests, produce incorrect policy answers, and reduce trust in reimbursement guidance, even though it does not directly execute code or access sensitive systems.

Content

No source excerpt is available for this finding.

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/线上费用报销流程管理办法V1.0.md (reported line 26)May include surrounding context.

md
目录

1 目的
\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\....
4

> 2 适用范围

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/线上费用报销流程管理办法V1.0.md (reported line 80)May include surrounding context.

md
目录

1 目的
\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\....
4

> 2 适用范围

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/线上费用报销流程管理办法V1.0.md (reported line 30)May include surrounding context.

md
4

> 2 适用范围
> \...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...
> 4
>
> 3 术语和定义

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/线上费用报销流程管理办法V1.0.md (reported line 62)May include surrounding context.

md
4

> 2 适用范围
> \...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...
> 4
>
> 3 术语和定义

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/线上费用报销流程管理办法V1.0.md (reported line 34)May include surrounding context.

md
> 4
>
> 3 术语和定义
> \...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\....
> 4
>
> 4 各部门职责

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/线上费用报销流程管理办法V1.0.md (reported line 38)May include surrounding context.

md
> 4
>
> 3 术语和定义
> \...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\...\....
> 4
>
> 4 各部门职责

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This section tells users to add account names, account numbers, bank names, and开户银行 information, which is sensitive financial data. The document provides procedural steps only and does not warn users about verifying the recipient account, limiting visibility appropriately, or handling this data carefully.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The file states that uploaded invoices will be automatically verified and mentions tax authority network synchronization, implying user financial document data is transmitted or checked against external systems. There is no accompanying notice about this data handling or privacy impact.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file contains the skill test report entirely in Chinese and does not state that the skill is Chinese-only, region-specific, or that users can choose another language. Under the policy, forcing a specific language without opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The contact information and project description are presented entirely in Chinese, with no indication that users can choose another language. Under the stated policy, language constraints should either be optional or clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The activation conditions do not define exclusions or boundary checks for adjacent topics such as general expense reimbursement, procurement, or entertainment expenses. This increases the chance of scope confusion, where users receive authoritative-sounding answers outside the intended policy domain, which is especially risky because the skill presents itself as based on official company documents.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file title and all user-facing instructional content are presented exclusively in Chinese, with no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

This markdown file is entirely written in Chinese and does not offer any language or locale choice. Under the stated policy criteria, forcing a specific language without user opt-in can be a natural-language policy violation unless the locale constraint is clearly documented and justified; no such justification is visible in the document body.

Content

No source excerpt is available for this finding.