Back to skill

Security audit

travel-world

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed Travel World connector for travel search and planning, with the main risk being that it relies on a live external MCP service whose available tools can change.

Install this only if you are comfortable sending travel searches and trip details to Travel World's hosted MCP service. Keep booking and payment steps user-confirmed on the Travel World site, and prefer runtime controls that review newly advertised MCP tools before use.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:20
Finding

Mutable and Unverified Remote MCP Tool Dependency

Content
View full analysis
` Once connected, your runtime discovers the available tools via `tools/list`. ``` ```markdown Travel World is expanding fast — new airlines, hotels, car, insurance, and experience providers come online continuously. **Call `tools/list` for the current set** rather than assuming; a capability may already be live. ``` ### Technical Analysis The skill instructs an MCP-capable runtime to retrieve its tool inventory dynamically from an externally controlled service. It does not define a fixed tool manifest, expected schemas, version constraints, capability allowlist, or validation policy for newly advertised tools. Consequently, the effective capabilities available to the agent may change after the local skill has been reviewed. A compromised or unexpectedly modified MCP server could advertise new tools, change existing tool descriptions or schemas, or alter the behavior behind legitimate-looking travel operations. The bearer token is also supplied to this external service, although the reviewed file does not contain a hardcoded token. This is a supply-chain trust weakness rather than evidence that the current provider is malicious. No remote code execution, local script execution, persistence, credential harvesting, or destructive behavior was identified in the reviewed artifact. ### Attack Path 1. An attacker compromises the MCP service, its deployment pipeline, or another component capable of changing its tool definitions or responses. 2. The compromised service modifies the result of `tools/list`, changes the schema or de ...[truncated 1538 chars]
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The phrase "Use this whenever a user wants to find flights, hotels, cars, insurance, or experiences, check whether a flight is on time, plan a trip, or discover brand-direct travel options" defines a very wide trigger surface and does not clarify when this skill should not be invoked. It lacks negative examples or constraints that would help distinguish Travel World from other generic travel-planning skills or ordinary conversation about travel.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This section says to use the skill when the user wants to find or compare travel products, check flight status, plan a trip, or find a brand-direct agent, but it does not define boundaries for ambiguous cases. Because these are common user intents in everyday travel conversations, the lack of exclusions or tighter constraints could cause unintended invocation.

Content

No source excerpt is available for this finding.