ClawHub
Back to skill

Security audit

pSEO Generator

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a pSEO guide, but it also includes brand-specific file loading and live Webflow publishing instructions without enough safeguards.

Review this skill before installing in any workspace with Tolstoy files or Webflow credentials. Use it only with explicit user approval before loading local business context or creating and publishing CMS items, and prefer draft or preview workflows first.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Session Persistence

Medium
Category
Rogue Agent
Content
- [ ] Define 20+ niches with full context (audience, pain points, monetization, subtopics)
- [ ] Choose 2-3 content types to start (resource pages recommended)
- [ ] Write strict TypeScript schemas with hard constraints
- [ ] Build generation script with niche injection and JSON validation
- [ ] Build React renderers per content type
- [ ] Generate first batch (start with 100 pages to validate quality)
Confidence
60% confidence
Finding
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The skill shifts from generic pSEO system design into tenant-specific operational instructions for gotolstoy.com, including local workspace paths, editorial directives, and publication workflow. That broadens the skill's authority and can cause an agent to access unrelated local business context or perform brand-specific actions not clearly implied by the general skill description.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill provides direct CMS publication steps for Webflow without any requirement for dry-run mode, environment scoping, or explicit user approval before modifying live content. An agent following these instructions could publish unwanted or harmful pages to a production site, causing integrity, reputational, and SEO damage.

Intent-Code Divergence

Low
Confidence
75% confidence
Finding
The skill establishes a strong architectural rule that renderers only consume pre-generated JSON and are kept separate from generation concerns. Later, the Webflow publishing flow has the renderer transform generated content into HTML/markdown for publishing, which is not a direct contradiction by itself, but it weakens the earlier 'only consume pre-generated JSON' claim by expanding renderer responsibilities into publishing output preparation.

Missing User Warnings

Low
Confidence
90% confidence
Finding
The instructions direct the agent to load multiple workspace-local files containing brand strategy, competitor analysis, keywords, and internal linking data, without warning that these may contain sensitive business information. This increases the chance of unnecessary data access and propagation of confidential context into later prompts or outputs.

Static analysis

No suspicious patterns detected.