Security audit

PeerBerry SDK

Security checks across malware telemetry and agentic risk

Overview

The skill matches its stated PeerBerry automation purpose, but it can guide real-money purchases and financial exports without consistently requiring explicit confirmation or secure data-handling warnings.

Install only if you intend to automate PeerBerry account workflows. Keep credentials, TOTP seeds, access tokens, exports, and logs private; start with read-only calls and DRY_RUN enabled; require a clear manual confirmation before any live purchase.

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 (4)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 80% confidence
Finding: The skill description frames the scope as investor automation and SDK usage, but the analysis indicates additional documentation-generation and code-introspection behavior not disclosed in the top-level description. Undisclosed behaviors increase trust risk because an agent may access source files, parse modules, or create docs artifacts beyond what a user expects, widening the operational surface and potentially exposing sensitive code or repository contents.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The quickstart repeatedly shows inline placeholders for email, password, and a TOTP/2FA secret but gives no guidance on secure secret handling, such as using environment variables, secret managers, or avoiding commits and logs. In an automation SDK for financial accounts, users often copy-paste examples into real projects, so unsafe secret-handling patterns can lead to credential leakage, account takeover, and exposure of investment data or trading capability.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: These recipes explicitly instruct the agent to perform real-money investment actions (`purchase_loan`) but do not require a prominent user confirmation or financial-risk warning at the point of execution. In an agent skill context, copy-paste prompts can normalize autonomous trading behavior and increase the chance of unintended purchases, especially for inexperienced users or when prompts are routed automatically.

Missing User Warnings

Low

Confidence: 83% confidence
Finding: The export recipe tells the user to save investments and transaction exports locally without warning that sensitive financial account data will be written to disk. In shared, synced, or insecure environments, this can expose private transaction history and portfolio information beyond the intended session.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.