ClawHub
Back to skill

Security audit

Mom AI Agent Parenting Knowledge

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a read-only parenting evidence helper, but it also bundles and links to an operational crawler playbook that can run scraping workflows and write article data.

Review before installing if you want a strictly read-only parenting helper. The normal KB instructions are coherent and safety-conscious, but the bundled crawler reference could lead an agent with matching local project access to run scraping and ingestion workflows. Prefer installing only after removing that playbook reference or confirming your agent environment cannot execute those operational steps.

SkillSpector

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

Description-Behavior Mismatch

High
Confidence
97% confidence
Finding
The playbook explicitly defines the skill as an operational crawl/search workflow rather than a read-only evidence retrieval behavior, which conflicts with the stated skill metadata and broadens the agent’s effective capabilities. In an agent environment, this kind of capability drift is dangerous because it can cause the model to perform data collection, site traversal, or ingestion-oriented actions that users did not authorize.

Description-Behavior Mismatch

Critical
Confidence
99% confidence
Finding
This section documents write-side behavior including inserting records into database tables and writing local JSON snapshots, directly contradicting the claim that the skill is strictly read-only. If an agent follows these instructions, it could mutate persistent state, create unauthorized datasets, or trigger data governance and operational integrity issues well outside the expected scope of a parenting evidence retrieval assistant.

Context-Inappropriate Capability

High
Confidence
98% confidence
Finding
The file includes executable crawler and browser-assisted scraping commands, including node scripts and Playwright usage, which are unrelated to a narrow read-only evidence retrieval role. Embedding such commands in a skill materially increases the risk that an agent will perform active scraping, browser automation, or broader reconnaissance instead of constrained retrieval.

Intent-Code Divergence

Medium
Confidence
86% confidence
Finding
Stating that the document is 'not a public evidence-answering spec' conflicts with the user-facing purpose of the skill and signals that the file was not intended to govern safe end-user interactions. In context, this mismatch increases the chance that internal operational instructions override user-safe behavior, causing the agent to act outside its advertised scope.

Static analysis

No suspicious patterns detected.