Security audit

Satellite Search

Security checks across malware telemetry and agentic risk

Overview

The skill is a disclosed satellite lookup tool, but it includes anti-bot scraping and automatic external query or LLM workflows that users should review before use.

Install only if you are comfortable with a tool that can scrape eoPortal using browser fingerprint stealth, refresh bulk local datasets, send fallback searches to DuckDuckGo, and send translation content to an OpenAI-compatible LLM provider when those commands run. Prefer local search/info commands for routine use, and avoid update/fetch/translate workflows in strict privacy, compliance, or website-terms environments.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The script contains explicit anti-bot evasion logic, including webdriver masking, fake Chrome/runtime properties, spoofed plugins/languages, and WebGL fingerprint manipulation. In the context of an offline satellite lookup skill, this exceeds normal data retrieval behavior and is dangerous because it is designed to bypass site defenses and access controls, creating legal, compliance, and platform-abuse risk if run against protected services.

Context-Inappropriate Capability

Low

Confidence: 80% confidence
Finding: The script exposes a shuffle option described as useful for evading rate limits on 'hot slugs,' indicating intent to alter request patterns to avoid defensive controls rather than simply fetch metadata. While lower severity than the stealth code, this still normalizes bypass behavior and can contribute to abusive scraping patterns against third-party services.

Missing User Warnings

Medium

Confidence: 83% confidence
Finding: The direct DuckDuckGo fallback sends user-provided search terms to a third-party service without any disclosure or consent mechanism in this code path. In a skill handling potentially sensitive satellite queries, this can leak user interests, internal project names, or proprietary terms to an external search provider and network intermediaries.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The translate subcommand delegates satellite descriptions to an LLM workflow without any explicit warning or consent boundary about sending content to an external model/provider. In an offline-first skill, this can mislead users into assuming processing stays local, causing unintended disclosure of indexed or fetched text to third-party services.

Ssd 3

Medium

Confidence: 86% confidence
Finding: The script injects untrusted source text and metadata directly into the LLM prompt, so malicious or crafted content in the eoPortal record could attempt prompt injection, override translation instructions, or induce the model to emit malformed or unintended output. In this script the impact is bounded because the result is parsed as JSON and written to a local dataset, but poisoned translations or prompt-following behavior could corrupt downstream data and any systems that later consume it.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.