ClawHub
Back to skill

Security audit

JSON Utils

Security checks across malware telemetry and agentic risk

Overview

This skill is a local JSON parsing and validation utility with no evidence of hidden network access, credential use, persistence, or destructive behavior.

Install only the documented Python dependencies from trusted package sources. Use the CLI on files you intend to read, and note that the batch processor can write a JSONL output file when you provide an output path. Treat the WebSearch/API-documentation section as optional external scripting-utils guidance, not part of this skill's installed capabilities.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
84% confidence
Finding
The documentation expands the skill into WebSearch and API-documentation workflows via a sibling skill path, which exceeds the manifest's JSON-focused purpose. Scope drift is dangerous because users may invoke broader capabilities than they intended, potentially introducing networked data flows or external content handling under the guise of a local JSON utility.

Description-Behavior Mismatch

Medium
Confidence
88% confidence
Finding
These examples actively advertise search-driven API documentation lookup, schema generation, and endpoint validation capabilities that are not described in the manifest. Such undocumented expansion can mislead operators about the trust boundary of the skill and cause unintended retrieval of remote content or use of additional tools outside the approved JSON-processing role.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal