ClawHub

领导模块

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed task-orchestration skill whose main risk is accidental interruption or metadata exposure from broad natural-language task commands.

Install only if you want your agent to coordinate background sub-agents and keep task state. Use explicit task-number commands when stopping or checking work, and avoid casual words like stop or interrupt during active multi-task sessions unless you intend to affect a task.

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 (3)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The interrupt commands are defined as common natural-language words ("stop" or "interrupt") without any explicit command prefix or interaction mode. In a multi-threaded agent architecture, this can cause accidental termination of background work when those words appear in ordinary conversation, quoted text, or task content, leading to unintended task cancellation and state inconsistency.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The status-check examples (e.g. "task list" / "what's running") are broad conversational phrases rather than clearly namespaced commands. That increases the chance that normal user dialogue, pasted logs, or adversarial content embedded in task data could be misinterpreted as a control request, exposing task metadata or altering agent behavior unexpectedly.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The documented interruption trigger uses very common conversational words like “停/stop” and “打断/interrupt,” which can be said incidentally in normal discussion rather than as an explicit control command. In a multi-threaded agent skill, that ambiguity can cause unintended task cancellation or denial of service against active background work, especially when multiple tasks are running and user utterances are short or context-poor.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal