ClawHub
Back to skill

Security audit

EU AI Act Compliance for Autonomous Agents

Security checks for vulnerabilities and agentic risk

Overview

The skill is a compliance guide, but it mixes sandbox assurances with production-ready code that can modify live GreenHelix agents, escrows, disputes, webhooks, and reports.

Treat this as production integration material, not a harmless sandbox tutorial. Before installing or using it, review every code example that calls GreenHelix, set an explicit sandbox URL for testing, use least-privilege test credentials, and avoid running escrow, dispute, webhook, or daemon examples against production until the operational and legal effects are approved.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Although presented as a guide, the file contains substantial operational code that performs live remote actions such as registering agents and services, creating escrows, opening disputes, publishing events, and registering webhooks. In a skill ecosystem, this is dangerous because users or automation may treat markdown examples as trusted documentation and execute them without realizing they modify external systems.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly frames itself as non-executable educational material, but later provides runnable production code that performs persistent monitoring, reads credentials, calls remote APIs, and writes files. This mismatch can mislead users or downstream agents into trusting and copying code they would not execute if the document were accurately labeled.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The document claims examples use a sandbox with no API key required, yet later states code runs against production endpoints and includes examples that read API credentials and operate on live services. That inconsistency increases the risk of accidental execution against real infrastructure, potentially causing unwanted registrations, payments, disputes, or data transmission.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The guide gives conflicting safety cues about sandbox versus production usage and does not provide a clear, consolidated warning that many examples target live APIs. This ambiguity can cause unsafe operational behavior, especially in a compliance-themed skill where users may assume the content is conservative and safe by design.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 1976)May include surrounding context.

md
---

*This guide provides technical compliance patterns for agent commerce systems. It is not legal advice. Consult qualified legal counsel for definitive guidance on EU AI Act compliance, Product Liability Directive obligations, and GDPR data processing requirements. All code examples use the GreenHelix A2A Commerce Gateway API at https://api.greenhelix.net/v1.*

---

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The skill references sensitive credentials, including an API key and signing key, without strong guidance on least privilege, storage, rotation, and avoiding exposure in copied examples. While the file does not directly exfiltrate secrets, it normalizes credential use in example code without adequate safeguards.

Content

No source excerpt is available for this finding.