Security audit

Epic Free Games

Security checks across malware telemetry and agentic risk

Overview

The skill is transparent about claiming Epic free games, but it automates account checkout with saved login cookies and includes Cloudflare anti-detection guidance.

Review carefully before installing. Use only if you accept the Epic account and terms-of-service risk, avoid the anti-detection instructions, protect or delete epic_auth.json, and do not enable cron unless you are comfortable with unattended account actions.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (8)

Context-Inappropriate Capability

Medium

Confidence: 98% confidence
Finding: These instructions advise use of `--disable-blink-features=AutomationControlled` specifically to reduce detection by Cloudflare, which is anti-bot evasion guidance rather than ordinary automation. That materially increases misuse potential by helping users circumvent a site's protective controls while automating account actions.

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The document first claims hCaptcha/Cloudflare cannot be bypassed automatically, but later provides techniques intended to avoid automation detection, creating a contradictory and misleading safety posture. This is dangerous because it may lower user caution while still embedding practical evasion steps for protected workflows.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The script saves authenticated browser state to `epic_auth.json`, which likely contains reusable session tokens or cookies for the Epic account. Storing that state on disk without explicit warning or restrictive file permissions increases the chance of local credential theft, accidental sharing, or misuse by other processes.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The script reloads persisted authentication state from disk and immediately reuses it for account actions. If the auth file is stolen, replaced, or unexpectedly shared, the script can operate under an account session without the user realizing stored credentials are being reused.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The script automates clicking the order/confirm button to claim content on the user's Epic account without an execution-time confirmation prompt. Although the purchase is intended for free items, it is still an account-affecting action and UI matching errors or page changes could cause unintended transactions or other irreversible actions.

Ssd 2

Medium

Confidence: 98% confidence
Finding: The phrasing frames disabling automation-detection signals as a normal setup step, but the practical effect is to evade website bot checks. Packaging such evasion as routine usage guidance normalizes bypass of service protections and increases the chance of abusive or policy-violating deployment.

Ssd 2

Medium

Confidence: 98% confidence
Finding: This troubleshooting section presents Cloudflare interception as a problem to work around and recommends the same anti-detection browser flag as the solution. That turns a defensive control into an obstacle to be bypassed, directly enabling evasion-oriented account automation.

Ssd 4

Medium

Confidence: 99% confidence
Finding: The full workflow combines anti-detection browser setup, page interaction, checkout automation, and credential-state persistence into an end-to-end evasion sequence for protected account actions. In context, this makes the skill substantially more dangerous because it operationalizes repeated automated purchases/claims against a service that deploys anti-bot controls.

VirusTotal

60/60 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.