ClawHub
Back to skill

Security audit

WP Multitool — WordPress Optimization Toolkit

Security checks for vulnerabilities and agentic risk

Overview

This WordPress optimization skill is mostly clear and purpose-aligned, but it includes a plugin activation command without clearly treating it as a state-changing action that needs user approval.

Review before installing if you run this on a production WordPress site. Read-only diagnostics are reasonably scoped, but require explicit approval before any cleanup, wp-config change, frontend toggle, database deletion, table optimization, or plugin activation.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

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 (3)

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill presents `wp plugin activate wp-multitool` in troubleshooting without any user-confirmation requirement, despite activation being a state-changing action that loads and executes plugin code. If an agent follows this guidance automatically, it could enable unreviewed functionality, alter site behavior, or trigger install-time hooks without operator consent. This is more dangerous in this skill because the rest of the document carefully distinguishes read-only from write actions, making the omission likely to mislead an automation agent.

Intent-Code Divergence

Low
Confidence
72% confidence
Finding
The file says all read commands in that section are safe to run without confirmation, but later documents `wp multitool redis --force` as re-querying the Redis server rather than merely reading cached local results. While still plausibly read-only, the comment overgeneralizes the behavior and conflicts with the later note that `--force` causes additional direct interaction with infrastructure.

Context-Inappropriate Capability

Low
Confidence
88% confidence
Finding
The troubleshooting section includes `wp plugin activate wp-multitool`, which changes site state by enabling code execution from a plugin. Although this is framed as operational guidance, it expands the skill beyond the declared read/cleanup/config scope and could cause an agent to perform a mutating action without explicit approval. In a WordPress environment, plugin activation can immediately alter behavior, run hooks, and trigger side effects.

Static analysis

No suspicious patterns detected.