Security audit

chrome-extension

Security checks across malware telemetry and agentic risk

Overview

This is a coherent Chrome extension development guide, but it includes under-scoped examples for bypassing or removing website security protections.

Use this skill only for authorized Chrome extension development. Before installing or following its output, review generated code for CSP removal, page script injection, broad host permissions, service-worker relays, persistent content scripts, and token storage; prefer narrow domain allowlists, runtime user consent, and avoiding CSP weakening unless there is a clearly justified defensive or enterprise need.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (8)

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The document presents an example manifest with an extension_pages CSP that appears relaxable while later text correctly states MV3 does not allow relaxing extension page CSP with unsafe-inline/unsafe-eval or remote scripts. In a reference skill, contradictory security guidance is dangerous because users may copy the invalid or insecure example and ship an extension with misunderstood CSP constraints, leading to unsafe workarounds or review failures.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The reference explicitly teaches removal of the `content-security-policy` response header to relax site protections. In a general Chrome extension development skill, this is a dangerous capability because it enables weakening page security boundaries and can facilitate script injection, data exfiltration, or integrity violations on visited sites.

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The dynamic-rule example shows runtime removal of `content-security-policy`, making it easy to operationalize page protection bypass on demand. This is especially risky because it converts a sensitive capability into copy-pasteable code without clear limitation, which could be repurposed to disable protections on arbitrary targets.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The skill explicitly advertises help for bypassing CSP from a content script and injecting scripts into web pages, but the skill metadata/description does not include a clear safety warning or boundary about legitimate, authorized use only. In an agent setting, that omission increases the chance the skill will be invoked for dual-use or abusive browser-extension behavior, especially because CSP bypass relays and page script injection are common primitives for data exfiltration, session abuse, and stealthy page manipulation.

Missing User Warnings

Medium

Confidence: 82% confidence
Finding: The document presents relay-based cross-origin fetching as the standard way to bypass page CSP from a content script, but does not prominently warn about privacy, trust-boundary, and abuse risks. Although the relay pattern itself can be legitimate, omitting risk guidance in a skill that encourages script injection and background communication lowers the barrier to misuse.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The CSP-removal examples omit a clear warning that disabling Content Security Policy weakens core browser-enforced defenses for script loading and network connections. Readers may treat this as routine extension engineering guidance, increasing the chance of unsafe deployment that undermines page integrity and user security.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The document states several permissions are 'safe to use freely' merely because they do not trigger an install warning, which can mislead developers into underestimating privacy and security risks. Permissions such as scripting, storage, dns, offscreen, unlimitedStorage, and enterprise-related APIs can still enable sensitive data access, code injection, persistence, or expanded attack surface depending on implementation.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The guidance explicitly recommends using extension storage for tokens and temporary auth data without any warning that Chrome extension storage is not an appropriate place for secrets in many threat models. In a Chrome extension context, content scripts and other extension components may access certain storage areas, and persisted tokens increase the blast radius of XSS, message-handling bugs, or extension compromise.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.