ClawHub
Back to skill

Security audit

Ollama集成工具

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a broad Ollama integration guide, but its local-only framing conflicts with API key and network guidance while requesting read, write, and command execution authority.

Review before installing. Use this only if you intentionally want an agent to run local Ollama-related commands and possibly manage files. Do not provide real API keys unless the publisher clarifies what service receives them, what endpoints are contacted, and when prompts or files leave the machine.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The skill claims to integrate a local Ollama runner, but the documentation also instructs users to configure an API key and describes API/network-based integration. This mismatch can cause users or orchestrators to grant unnecessary network access or secrets to a skill they believe is purely local, increasing the chance of unintended data exposure or over-privileged execution.

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The file states the skill is 'local-only' while simultaneously requiring API key setup and discussing HTTPS/network protections. Conflicting trust signals are dangerous because operators may approve the skill under a lower-risk local-data assumption while it actually depends on remote connectivity and secret material.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The activation guidance is overly broad, covering generic AI model invocation, conversation, agent orchestration, and LLM applications. In agent ecosystems, vague triggers can cause the skill to be selected for many unrelated tasks, which is risky here because the skill exposes exec/write capabilities and may be invoked in contexts where those powers are unnecessary.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.