ClawHub
Back to skill

Security audit

A2WF — Agent-to-Web Framework

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward A2WF siteai.json validator/generator with disclosed command and optional website-fetch use, though its live-audit documentation is somewhat inconsistent.

Before installing, understand that normal local validation is offline, but live website checks may use web_fetch to retrieve a site's siteai.json. The audit command documentation is not fully aligned with the script implementation, so treat URL audit examples as guidance rather than a fully implemented network-auditing feature.

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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (2)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The documentation tells users the validator is 'local-only' and makes 'no network requests,' but elsewhere instructs live website fetching and shows an audit mode against a URL. This inconsistency can mislead operators into treating the skill as offline-safe, causing it to be invoked in more trusted contexts while still enabling external network access and URL-driven behavior.

Content

No source excerpt is available for this finding.