Security audit

live-stream-monitor

Security checks for vulnerabilities and agentic risk

Overview

This skill has a coherent live-stream monitoring purpose, but it needs Review because it can use microphone-backed browser speech recognition without clear privacy boundaries or user consent controls.

Review before installing. Use it only if you are comfortable with a browser session that may access your microphone while a stream plays through speakers, and avoid running it around private conversations. The publisher should add explicit consent steps, explain the audio and transcript data flow, remove or justify media auto-approval, and either implement the advertised keyword alerts or narrow the claims.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 98% confidence
Finding: The skill claims to monitor stream audio, transcribe speech, detect keywords, and notify users, but the documented behavior does not define the actual mechanisms or boundaries and appears to require sensitive browser/media capabilities not transparently declared. That mismatch is dangerous because users may grant microphone/browser permissions under incomplete pretenses, enabling unexpected audio capture or browser automation behavior beyond what the skill clearly explains.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The README instructs users to grant microphone access and use browser speech transcription, but it does not disclose the privacy and consent implications of capturing and processing audio during live stream monitoring. Even if the source is a public stream, the use of browser or cloud-backed speech recognition may transmit audio or transcripts to third-party services, creating privacy, legal, and user-trust risks.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The trigger language is broad enough that the skill could activate in situations outside a user's clear intent, especially for generic requests about monitoring, transcription, or alerts. Over-broad activation increases the chance of unnecessary collection of stream or ambient audio and can cause a sensitive-capability skill to run when a safer or narrower tool should have been used.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: Requesting microphone access without a clear privacy warning is a real issue because microphone permission can expose ambient conversations, not just intended stream audio. In this context, the danger is heightened because the skill says it monitors online streams, yet the stated requirement implies local audio capture, creating ambiguity about whether user/environment audio may be recorded or processed.

Natural-Language Policy Violations

Medium

Confidence: 95% confidence
Finding: The code hard-codes `recognition.lang = 'en-US'`, which imposes a specific language/locale for transcription. This matches the policy category for language or locale constraints because the skill does not offer opt-in, selection, or explain why only English (US) is appropriate.

Static analysis

No suspicious patterns detected.