tinyhumansai OpenHuman

• CVE-2026-49407 High 7.7 OpenHuman: Unauthenticated dictation WebSocket allows arbitrary audio streaming and unbounded memory growth 0.3%
• CVE-2026-54142 High 8.7 OpenHuman: Indirect prompt injection via background email sync executes Composio actions without user approval 0.5%
• CVE-2026-54146 High 8.8 OpenHuman: Bearer token leak via process environment variable on macOS (same-UID local RCE primitive) 0.2%
• CVE-2026-54143 Critical 9.8 OpenHuman allows remote inbound channel messages to create unapproved shell cron jobs leading to host command execution 1.2%
• CVE-2026-54141 Critical 9.0 OpenHuman: Yuanbao non-web channel can bypass human approval to execute `node_exec` 0.8%
• CVE-2026-49393 High 8.4 OpenHuman: Unauthenticated Socket.IO RPC bypass allows invoking protected OpenHuman core controllers without the Bearer token 0.3%
• CVE-2026-54145 High 8.8 OpenHuman: Bearer token leaked via approval gate session_id (SQLite plaintext + debug log + RPC response) <0.1%
• CVE-2026-49405 Medium 6.2 OpenHuman: Unauthenticated server-sent events endpoint leaks user-facing event data 0.3%
• CVE-2026-54144 Medium 6.1 OpenHuman: Embedded Google Meet webview can silently capture the macOS desktop without a user gesture 0.5%
• CVE-2026-54140 High 8.3 OpenHuman: MCP tool description injection: untrusted server descriptions execute in agent LLM context without sanitization 0.6%
• CVE-2026-49410 High 8.1 OpenHuman /auth/telegram CSRF allows attacker-controlled session implantation 0.2%
• CVE-2026-55743 Critical 9.6 OpenHuman desktop agent shell tool sandbox bypass leads to arbitrary command execution 0.4%