Back to skill

Security audit

spidey-weather

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a weather skill with no confirmed malicious behavior, so the verdict is benign, but the review is incomplete and flags a mismatch between the described city support and the reported behavior.

The verdict is benign, but confidence is medium and the scan is partial: only 3 of 8 components were fully inspected, and tool-misuse analysis was degraded. Before installing, consider that the supplied command installs an external package whose behavior could not be checked, and the reported implementation may not honor requests for cities other than Los Angeles.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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 (4)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The only supplied command performs package installation, a materially different operation from the declared user-facing weather functionality. The package name suggests a relationship to the description, but its implementation is not supplied, so its behavior cannot be verified. This assessment does not imply that the installed package is malicious.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · lib/weather.mjs (reported line 4)May include surrounding context.

js
import { weatherIcon } from "./weather-presentation.mjs";

const WEATHER_URL = new URL(
  "https://api.open-meteo.com/v1/forecast"
);

const TIMEZONE = "America/Los_Angeles";

The weather collector always requests Los Angeles weather despite the manifest claiming support for a user-supplied city.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

The manifest describes Los Angeles as a default and promises support for a city supplied in the request. However, collectWeather accepts only fetcher and now, fetches the endpoint with coordinates hardcoded at L045–L046, and normalizes the result using the default Los Angeles location and timezone at L328.

Content

No source excerpt is available for this finding.

Geocoding results are forced to English without a user language choice.

Low
Category
Output Handling
Confidence
90% confidence
Finding

The geocoding request unconditionally sets language=en, causing localized place names used in user-facing labels to be returned in English. No language opt-in or documented justification for this restriction is provided.

Content

No source excerpt is available for this finding.