ClawHub
Back to skill

Security audit

本地知识库集成

Security checks for vulnerabilities and agentic risk

Overview

This local knowledge-base skill has no executable payload, but it asks for broad read/write/command authority with unclear limits and inconsistent local-versus-network guidance.

Install only if you are comfortable giving the skill local file read/write and command-execution capability. Before using it with sensitive documents, require explicit confirmation for document ingestion, API use, network calls, and any shell command execution.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
The skill is presented as a local knowledge-base/document retrieval tool, yet it requests broad `exec` capability alongside generic automation language. That mismatch expands the attack surface significantly because a user or downstream agent could be induced to run arbitrary system commands unrelated to document retrieval, enabling file, process, or environment access beyond the stated purpose.

Intent-Code Divergence

Medium
Confidence
84% confidence
Finding
The documentation states the skill is local-only, but elsewhere advertises API integration and network communication. This inconsistency is dangerous because operators may trust the skill with sensitive local documents under the assumption that data never leaves the machine, when in fact external transmission may occur.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The activation guidance is extremely broad, describing use in general productivity, automation, and workflow optimization contexts. Such vague triggering criteria can cause the skill to activate in many unrelated situations, increasing the chance that powerful tools like read/write/exec are invoked when a narrower, safer skill should have been used.

Vague Triggers

Medium
Confidence
82% confidence
Finding
The example command phrase is vague and underspecified, making it a poor safety boundary for skill activation. Ambiguous triggers can lead an agent to interpret ordinary conversational text as authorization to access documents or switch modes, which is especially risky in a skill that also has filesystem and execution capabilities.

Static analysis

No suspicious patterns detected.