Security audit

Canvas Json Handler Free

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a JSON Canvas editing guide, but it asks for broad read/write/exec authority while advertising itself for unrelated analytics and automation tasks.

Install only if you intend to use it for JSON Canvas files, and avoid allowing it to auto-run for general analysis or reporting requests. Review any proposed file writes, callbacks, or network diagnostic commands before approving them.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Behavioral ASTexec() Call, eval() Call, Dynamic Import
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The manifest advertises a narrow JSON Canvas file handler but also claims applicability to broad data analysis, reporting, and statistical insight tasks. This can cause the skill to be invoked in unrelated contexts, leading agents to apply file read/write and exec-capable behavior where users did not intend a canvas tool, increasing the chance of unsafe actions or misleading outputs.

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: A local JSON Canvas processing skill has no clear need for callback URLs, ping-based network diagnostics, or similar exec-driven network behavior. These capabilities expand the attack surface by enabling outbound connectivity and shell execution patterns that could be repurposed for exfiltration, environment probing, or unintended command execution.

Vague Triggers

High

Confidence: 95% confidence
Finding: The description uses an overly broad trigger scope far outside the stated domain of JSON Canvas handling. This creates skill-routing risk: the agent may select a file-writing, exec-capable skill for ordinary data-analysis requests, unnecessarily exposing local files and execution features in contexts where they are not needed.

Vague Triggers

High

Confidence: 96% confidence
Finding: The capability coverage section intentionally enumerates broad everyday keywords to maximize matching, including generic terms like automation, analysis, and visualization. In an agent ecosystem, this acts like prompt-scope overreach and can hijack unrelated requests, causing an overprivileged skill with read/write/exec tools to run where a safer, narrower skill should have been chosen.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The skill instructs the agent to write files and use exec-based operations without clearly warning that local files may be modified. In practice, this can lead to silent or surprising filesystem changes, especially if the skill is auto-selected through its overly broad scope.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.