ClawHub
Back to skill

Security audit

Prompt Refiner

Security checks across malware telemetry and agentic risk

Overview

This appears to be a prompt-rewriting skill with documentation hygiene issues, not a skill that itself accesses accounts or sends data externally.

Install only if you are comfortable with a text-only prompt helper. Treat any Gmail, Slack, or other account-related examples as templates only: replace personal identifiers with placeholders, do not paste private message content unnecessarily, and confirm before posting or sending anything externally.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Context-Inappropriate Capability

Low
Confidence
92% confidence
Finding
The documentation includes real-looking sensitive context such as a personal Gmail address and an absolute local filesystem path, which can normalize disclosure of personal or environment-specific data and may expose private information if copied into prompts or logs. Even in examples, such details can be harvested, reused in social engineering, or inadvertently propagated into downstream systems.

Missing User Warnings

Low
Confidence
89% confidence
Finding
The Slack-posting example normalizes sending generated content to an external service without any consent, review, or data-sensitivity check. This can encourage users or downstream agents to treat external transmission as routine, increasing the risk of accidental disclosure of internal or sensitive information.

Missing User Warnings

Low
Confidence
91% confidence
Finding
The Gmail example frames access to a personal inbox as a normal refinement target without warning about the sensitivity of email data or the need for explicit authorization. This can desensitize users to privacy boundaries and encourage prompts that expose account details, message content, and other personal data.

Ssd 3

Medium
Confidence
95% confidence
Finding
Including a real-looking personal Gmail address in documentation unnecessarily embeds identifiable account data into the skill artifact and normalizes placing personal account identifiers inside prompts. This creates privacy risk, potential targeting value, and the chance that users will imitate the pattern with real credentials or accounts.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.