Obscura Skill

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only browser automation skill, but it needs Review because it broadly steers agents toward stealth anti-bot scraping without clear authorization checks.

Review before installing if you work with third-party websites or authenticated accounts. Only use stealth or anti-detection features where you have explicit permission and policy/legal clearance, avoid production credentials, keep any CDP server bound to localhost, stop it after use, and prefer pinned or verified Obscura binaries installed with the least privilege needed.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 97% confidence
Finding: The skill explicitly recommends using stealth and anti-detection to 'defeat trivial bot-detection' and for bot-protected pages. That goes beyond ordinary browser automation or E2E testing and materially facilitates concealment of automated access to third-party sites, increasing abuse and policy-evasion risk.

Context-Inappropriate Capability

High

Confidence: 99% confidence
Finding: The Stealth Mode section documents detailed fingerprint spoofing and automation-concealment features such as masking webdriver, native-function masking, and per-session fingerprint randomization. These capabilities are designed to evade detection rather than merely render pages, making the skill substantially more dangerous in a scraping context.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The skill advertises very broad automatic trigger conditions spanning generic topics like web scraping, headless browsers, JS rendering, CDP, and anti-bot behavior. In an agent environment, this can cause the skill to activate in contexts the user did not intend, introducing stealth-scraping guidance or browser automation behavior into unrelated tasks and increasing the chance of unsafe or policy-sensitive actions.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The README promotes 'enable stealth mode' and anti-detection usage without any warning, limitation, or compliance guidance. Because this skill is specifically for scraping and headless automation, omission of a warning makes the capability more dangerous by normalizing techniques commonly used to evade site protections, which can facilitate unauthorized scraping or abuse.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill recommends stealth for bot-protected pages without front-loading meaningful warnings about terms-of-service violations, access restrictions, or legal and ethical consequences. Presenting evasion guidance as routine usage normalizes risky behavior and may lead agents to assist improper scraping by default.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal