Banks Get OpenClaw Security Risk Alerts

🔑 Enhanced Key Takeaways

• •A critical vulnerability (CVE-2026-28450) in OpenClaw versions prior to 2026.2.12 allows unauthenticated attackers to bypass authentication via exposed Nostr plugin endpoints, enabling profile data exfiltration and malicious event publishing[1][2].
• •Security audits revealed high rates of vulnerabilities in OpenClaw skills: Snyk found 13% critical flaws across 3,984 ClawHub skills, while Cisco identified 26% vulnerable in 31,000 agent skills including command injection and data exfiltration[1].
• •OpenClaw saw explosive growth post-launch in November 2025, reaching 100,000+ GitHub stars by late January 2026, with 135,000 exposed instances detected globally and 22% of enterprises deploying it as shadow IT without IT approval[5].
• •A 1-click remote code execution flaw (CVE-2026-25253) enables token theft and arbitrary command execution via malicious links, while a Moltbook database misconfiguration exposed 1.5 million API tokens and 35,000 emails[1][4].

🛠️ Technical Deep Dive

• •CVE-2026-28450 (CWE-306 Missing Authentication) affects Nostr plugin endpoints /api/channels/nostr/:accountId/profile and /api/channels/nostr/:accountId/profile/import, allowing unauthenticated HTTP GET/POST requests to read/modify profiles and sign events with bot private keys[2].
• •Attack chain for hijacking: malicious link triggers Control UI to connect to attacker server, leaking auth token for full instance control including data reading, command execution, and agent modification[1].
• •Skills ecosystem risks: malicious payloads in 76/3,984 ClawHub skills for credential theft/backdoors; 26% of 31,000 skills vulnerable to prompt injection, command injection, data exfiltration[1].
• •Moltbook exposure: misconfigured Supabase database revealed in client-side JS, leaking 1.5M API tokens, 35K emails, private agent messages[1].

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline