Security audit

Encoding Toolkit Free

Security checks across malware telemetry and agentic risk

Overview

This is a local encoding and decoding helper with some overbroad wording, but I found no hidden persistence, credential use, exfiltration, or destructive behavior.

Install only if you want an exec-capable helper for local encoding, decoding, JWT inspection, and checksum work. Treat JWTs and files as potentially sensitive, prefer local operations, and do not use online hash or external API services unless you explicitly intend to share that input.

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)

Description-Behavior Mismatch

Medium

Confidence: 88% confidence
Finding: The skill is presented as a local encoding/decoding toolkit, but the dependency and API-key sections broaden behavior to include external APIs/services without clear scope controls. That scope drift can cause an agent to access network resources or process data outside user expectations, increasing the chance of unintended data disclosure or over-privileged execution.

Context-Inappropriate Capability

Low

Confidence: 83% confidence
Finding: The error-handling section instructs generic network diagnostics such as ping testing, which is unrelated to the stated encoding/decoding purpose. Even limited network diagnostic guidance can nudge an agent into unnecessary outbound activity, creating avoidable exposure and expanding the operational scope beyond least privilege.

Intent-Code Divergence

Medium

Confidence: 79% confidence
Finding: The skill describes broad create/query/export/import/convert operations that imply more capability than the documented six encoding scenarios and later free-version limits. This ambiguity can cause an agent to apply the skill in unintended contexts, potentially handling broader data transformations or file operations than users expect.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The activation guidance is overly broad, including generic code generation, programming assistance, testing, and deployment use cases rather than narrowly scoped encoding tasks. Overbroad invocation criteria increase the likelihood that an agent will select this skill in inappropriate contexts where exec-enabled behavior may be unnecessary or risky.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The catch-all keyword coverage list is highly ambiguous and includes fragmented, generic terms that could match many unrelated requests. In an exec-capable skill, overly permissive triggering increases the chance of accidental activation and misuse outside the intended security and functional boundaries.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.