ClawHub

Eli Prompt Guard

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed prompt-injection guard that scans outbound automated submissions and pauses risky ones for owner review.

Install this if you want automated outbound posts, forms, API calls, or shared file writes checked before submission. Confirm who the configured Owner is, whether full-content review is acceptable for your data, and keep platform and logging settings scoped to the workflows you actually use.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
85% confidence
Finding
The catch-all trigger 'Any automated task that involves external submission' is broad enough to fire in ambiguous contexts, which can cause inconsistent enforcement or guard bypass if integrators interpret it differently. In a protection skill, vague activation criteria reduce reliability and may leave risky submission paths unchecked.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The trigger 'Executing commands that include user-provided content' is underspecified and does not clearly define what counts as a command, which execution environments are covered, or whether indirect/template-expanded content is included. That ambiguity can lead to partial deployment and missed checks before dangerous command execution flows.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal