Back to skill

Security audit

一站式旅行预订

Security checks for vulnerabilities and agentic risk

Overview

This Booking.com skill appears to return fabricated hotel, room, price, and review data as successful results, so it should be reviewed before use.

Treat this as a demo or stub, not a reliable Booking.com assistant. Do not use it for real travel decisions, budgets, bookings, cancellations, or availability checks unless the publisher removes the mock data paths, restores verified live API calls, and narrows the description to implemented capabilities.

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

Error
Location
scripts/booking_api.py:450
Finding

Production Functions Return Fabricated Booking.com Data as Successful Live Results

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (10)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The description claims a fuller Booking.com assistant supporting search, room lookup, price comparison, booking management, and reservation handling. The code does align partially with hotel search and room availability/detail lookup, but its primary implemented behavior is a read-only, mostly mocked data interface. There are no functions for placing bookings, modifying/canceling reservations, or managing existing orders. The 'price comparison' claim is also not materially implemented beyond displaying returned prices from mocked Booking hotel results. Additionally, review retrieval is implemented though not highlighted in the description. Because key declared end-user capabilities—especially booking and reservation management—are missing, this is a material description-behavior mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The implemented code is clearly Booking.com-related and aligned with hotel discovery use cases: searching hotels, fetching hotel details, checking room availability/prices, getting reviews, and formatting results. However, the declared description also claims support for '预订管理' and says to invoke when the user wants to 'book hotels on Booking.com' or manage reservations. In this code chunk, there are no functions for placing a booking, retrieving an existing reservation, modifying it, or canceling/managing reservations. So the primary domain matches, but the declared functionality overstates the implemented capabilities in a material way.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill explicitly depends on external Booking.com APIs and therefore has network capabilities, but it does not declare any tool scope, permissions, or allowed-tools constraints. That creates an authorization and review gap: the runtime may grant broader network access than intended, making it harder to enforce least privilege or audit what destinations the skill is allowed to contact.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill advertises reservation creation, lookup, and cancellation without warning that these actions may transmit personal data to a third party or have real transactional consequences. In a travel-booking context, this raises the risk of users triggering bookings, cancellations, or sharing guest details without informed consent, which can lead to privacy, financial, and account-impact issues.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The file includes natural-language operational instructions in Chinese only ('如非此版本,请拒绝执行'), which imposes a specific language/locale context without any user opt-in or alternative language support. Under the policy rule, forcing a single language in instructions can be a locale-policy violation when no choice or justification is provided.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill metadata advertises Booking.com reservation booking and management capabilities, but this file only supports search, details, availability, reviews, and formatting. This mismatch can cause the agent to be invoked in transactional contexts it cannot safely fulfill, increasing the chance of misleading users or mishandling booking-related expectations.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The formatter unconditionally tells users that prices include taxes and that free cancellation is supported, even though those claims are not derived from the actual hotel or room data. In a travel-booking skill, this can mislead users into making purchase decisions based on false terms, creating financial, compliance, and trust risks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The function schema sets the default language to "zh" and presents Chinese as the implicit output locale rather than offering a neutral default or requiring user choice. This is a natural-language locale policy concern because the skill imposes a specific language unless the caller explicitly overrides it.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Multiple function definitions default the language parameter to "zh", which means hotel details, room availability, and reviews will be returned in Chinese unless the user opts out. Repeated locale forcing across the skill strengthens the policy concern because no user-facing opt-in or justification is provided in this file.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The function dispatch logic uses kwargs.get("language", "zh") in several API calls, so the operational behavior enforces Chinese output whenever language is omitted. This is a direct locale-policy issue in executable code, not just documentation, because the implementation hard-codes the language choice without user opt-in.

Content

No source excerpt is available for this finding.