Back to skill

Security audit

Lounge Guru Map

Security checks for vulnerabilities and agentic risk

Overview

This is a mostly coherent offline airport catalog skill, but its npm installation path is not reproducibly pinned and the referenced dependency file is missing.

Review the install path before using this skill. The runtime itself appears local and read-only, but the current bundle can require npm to resolve mutable dependencies from the registry; install it only in a constrained environment or after the publisher supplies exact pinned dependency metadata and a lockfile. Also expect results to include airport-adjacent dining, spa, rest, and work-pod records, not only traditional lounges.

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:9
Finding

Unpinned and Unverifiable npm Runtime Dependencies

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:9-14 and SKILL.md:40-42
Vulnerability Type: Unpinned third-party dependencies and missing dependency integrity metadata
Risk Level: Medium

Vulnerable Code

yaml
"install":
  [
    {
      "id": "node",
      "kind": "node",
      "package": "@modelcontextprotocol/sdk",
      "label": "Install MCP runtime dependencies (run npm install in bundle root)",
    },
  ],
markdown
- Install package dependencies in the bundle root before running runtime scripts:
  - `npm install`
- Required packages are declared in `package.json` (`@modelcontextprotocol/sdk`, `zod`).

The audited project does not contain the referenced package.json or an npm lockfile. The installer metadata also identifies @modelcontextprotocol/sdk without an exact version, while scripts/runtime/contract.mjs directly imports the additional zod dependency.

Technical Analysis

The package cannot establish which dependency versions and transitive components were reviewed. An unpinned package reference can resolve to a different release as the registry changes, and the absence of a lockfile prevents integrity-controlled, reproducible installation.

This is a supply-chain weakness rather than evidence that the currently named package is malicious. However, installing mutable dependencies allows a compromised future package release, compromised transitive dependency, or registry-account takeover to introduce code that was not part of this audit.

npm packages may execute lifecycle scripts during installation. Such scripts run with the privileges of the user or automation process performing the installation. The runtime also executes the installed SDK when the MCP server starts.

Attack Path

  1. An operator installs the Skill and follows its dependency installation instructions, or the marketplace installer processes the unversioned package field.
  2. npm resolves the package and its transitive dependencies us ...[truncated 1405 chars]
Remediation
View remediation

Remediation Suggestions

  1. Include a minimal, reviewed package.json declaring exact versions of both @modelcontextprotocol/sdk and zod.
  2. Include the corresponding package-lock.json in the published bundle.
  3. Replace npm install with npm ci so installation uses the exact lockfile graph and fails when metadata is inconsistent.
  4. Pin the installer metadata to an exact reviewed package version rather than allowing mutable latest-version resolution.
  5. Review transitive dependencies and npm lifecycle scripts before publication.
  6. Use a trusted registry and retain npm integrity hashes in the lockfile.
  7. Consider installing with lifecycle scripts disabled where compatible:
bash
npm ci --ignore-scripts
  1. Run the MCP server as an unprivileged user in an environment with outbound network access disabled, preserving the declared offline-only boundary even if a dependency is compromised.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 28)May include surrounding context.

md
- Offline-only runtime: use bundled `assets/catalog.json` via local stdio MCP. Treat it as a general airport-lounge catalog, not a single-program entitlement so

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest frames this skill as an offline airport lounge catalog, but the bundled data clearly includes non-lounge venue types such as EAT, REFRESH, REST, and UNWIND in the top-level filters and throughout the records. That means the actual behavior supported by this asset extends beyond lounge lookup into restaurants, spas, sleep pods, and gaming/work pods, which is broader than the stated lounge-focused scope.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The dataset supports discovery of bars, restaurants, spas, sleep suites, gaming booths, and work pods, not just lounges. While airport-adjacent, these capabilities are a context expansion beyond the manifest's stated purpose of offline airport lounge lookup, filtering, briefs, and comparisons.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

This JSON repeatedly embeds URLs under the fixed path segment "en-GB", which is a natural-language locale choice rather than a technical necessity apparent from the file. Because the file provides no indication that this UK-English locale is optional, region-specific, or user-selectable, it can conflict with an organizational policy requiring language or locale choice.

Content

No source excerpt is available for this finding.