ClawHub
Back to skill

Security audit

agent-builder-tool

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed AgentPMT integration for creating and managing remote AI agents, with no hidden local runtime or deceptive behavior found.

Install this only if you want ordinary agent-building requests to use AgentPMT. Before state-changing actions such as publish, archive, remix, or replacing context_document_ids, confirm the target agent and intended changes; keep AgentPMT credentials in the setup flow and out of prompts or logs.

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
95% confidence
Finding
The top-level skill description contains very broad activation language such as 'build a custom AI agent' and 'turn a job description into a working ai agent,' which can match many ordinary user requests unrelated to this exact product integration. Over-broad routing can cause the agent to invoke this skill unexpectedly, sending user content into remote AgentPMT tool flows and increasing the chance of unintended external actions or data disclosure.

Vague Triggers

Medium
Confidence
97% confidence
Finding
The 'Search and activation keywords' section lists generic phrases that are common in ordinary AI-assistant conversations and lacks limiting conditions. This increases the risk of accidental invocation and privilege overreach, especially because the skill can create, publish, and modify remote agents through authenticated API calls.

Missing User Warnings

Medium
Confidence
81% confidence
Finding
The schema states that `context_document_ids` on `create_new`/`update_existing` replaces the full existing set, but it does not prominently warn that callers can unintentionally remove previously attached context documents. In an agent-building tool, this can silently strip grounding, policy, or business context from an agent and materially change its behavior or data exposure boundaries.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.