Security audit

图谱

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a blockchain data-routing helper, but it requests broad read, write, and command execution authority without clear boundaries.

Review this before installing. It may be useful for blockchain data questions, but grant it only the minimum tools needed, avoid giving write or shell execution access unless you have a specific reason, and use a limited API key because the documented behavior and output contract are inconsistent.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The skill documentation contradicts itself about whether it is pure Markdown guidance or a skill that can execute commands. This ambiguity is dangerous because an agent or user may authorize command execution under the assumption that the skill is non-executing, increasing the chance of unexpected privileged actions.

Intent-Code Divergence

Medium

Confidence: 84% confidence
Finding: The documented return schema appears to describe a generic scoring/audit result rather than blockchain routing or graph-query output. This mismatch can mislead downstream automation into trusting incorrect data shapes, causing unsafe chaining, bad routing decisions, or improper handling of responses.

Context-Inappropriate Capability

High

Confidence: 96% confidence
Finding: The skill declares read, write, and exec capabilities even though its stated purpose is only routing blockchain data questions to Graph Protocol services. Excess capabilities violate least privilege and could allow file modification or shell command execution unrelated to the task, especially if an agent grants tools based on the manifest.

Vague Triggers

Medium

Confidence: 81% confidence
Finding: The trigger language is broad enough that many generic data-analysis or reporting requests could invoke this skill unintentionally. Over-broad activation is risky because it can expose external services, credentials, or elevated tools in contexts where the user did not intend to use this integration.

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The examples and capability framing cover many endpoints and analytical tasks without defining clear boundaries for when the skill should activate. In an agent ecosystem, this can cause tool over-selection and broaden the situations in which external calls or privileged operations are attempted.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.