ClawHub

Prompt Safe

Security checks across malware telemetry and agentic risk

Overview

This is a local prompt-building helper with some overstated safety and privacy documentation, but no hidden access, exfiltration, or destructive behavior was found.

Install only if you are comfortable using a heuristic prompt assembly helper. Test its token estimates against your actual model and tokenizer, and ensure any long-term memory store has user consent, review, deletion, and sensitive-data limits.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The skill claims to guarantee prevention of token-overflow API failures, but the implementation relies on rough character-based heuristics and only checks a computed estimate against a safety threshold rather than the real tokenizer or actual provider limits. In a prompt-assembly component, this mismatch can cause oversized prompts to be sent despite the advertised guarantee, leading to request failures, dropped context, or reliability issues that may cascade into agent malfunction.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The document explicitly allows storing user identity and communication preferences in long-term memory, but it provides no requirement for informed user consent, data minimization beyond token cost, retention limits, or sensitivity handling. In a memory-orchestration skill, this increases the chance that personally identifiable information is persistently retained and reused across sessions in ways the user may not expect, creating privacy and compliance risk.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal