Security audit

Pangolinfo Amazon Niche Finder

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a coherent Amazon category analytics integration, but it explicitly tells the AI to read and reason about a raw Pangolinfo API key from the environment.

Install only if you are comfortable with the agent being instructed to access the Pangolinfo API key in its environment. Prefer configuring credentials in the MCP server or a managed secret store, and avoid asking the agent to print, validate, transform, or troubleshoot the raw key value.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (2)

Context-Inappropriate Capability

Medium

Confidence: 86% confidence
Finding: The skill instructs the AI to directly read authentication material from environment variables and reason about key validity/format. In an adversarial or over-permissive agent environment, this increases the chance of secret exposure to the model layer or accidental disclosure in logs, traces, tool arguments, or user-visible output, which is unnecessary for a category analytics workflow.

Ssd 3

Medium

Confidence: 90% confidence
Finding: The instructions normalize direct secret handling by the AI, including reading environment/config sources and recognizing specific key formats. Even without explicit exfiltration text, this broadens secret exposure to the model execution context and can make prompt-injection or accidental leakage more damaging than necessary.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.