2026年6月5日の
セキュリティ情報

Subject: [Security Advisory] Supply Chain Risks in AI Agent "Skills"

Dear IT/Security Team,

We are sharing information regarding a bypass in the detection mechanisms for AI skills on ClawHub, Cisco, and Vercel platforms.

■ Overview
Research by Trail of Bits indicates that malicious AI skills can bypass security scanners using techniques such as excessive newline insertion or hiding payloads in Python bytecode (.pyc) and archive files. This allows attackers to distribute malicious code via public marketplaces to perform actions like environment variable exfiltration.

■ Scope
- Environments utilizing AI skills from ClawHub, Cisco, and Vercel marketplaces.

■ Recommended Actions
1. Audit and suspend the use of unverified third-party skills from public marketplaces.
2. Implement supply chain security best practices, including the use of controlled repositories and version pinning for all AI components.
3. Restrict the execution of AI agents in high-privilege or sensitive environments.

■ Reference
- Trail of Bits Research Report

Priority: Medium
Deadline: As soon as possible

Subject: [Security Alert] Magecart Campaign Abusing Stripe API for Data Exfiltration

Dear IT/Security Team,

We are sharing information regarding a new Magecart campaign targeting e-commerce checkout pages.

■ Overview
Attackers are using Google Tag Manager (GTM) to inject skimmers that exfiltrate stolen credit card data via api.stripe.com. By leveraging trusted domains, the malware bypasses Content Security Policy (CSP) and network filters.

■ Scope
- E-commerce sites utilizing GTM and Stripe for payment processing.

■ Action Plan
1. Audit GTM containers for any unauthorized tags or suspicious custom HTML scripts.
2. Monitor network traffic to api.stripe.com to identify anomalous data exfiltration patterns that deviate from standard payment flows.
3. Review and tighten GTM access permissions to prevent unauthorized modifications.

■ Reference
- Sansec Security Research

Priority: High
Deadline: Immediate