{"skill":{"slug":"agent-guard","displayName":"Agent Guard","summary":"Real-time prompt injection and command injection detection for OpenClaw agents. Screens incoming messages, tool results, GitHub issues, and external content...","tags":{"latest":"1.0.1"},"stats":{"comments":0,"downloads":473,"installsAllTime":1,"installsCurrent":1,"stars":0,"versions":2},"createdAt":1773342770214,"updatedAt":1778491865024},"latestVersion":{"version":"1.0.1","createdAt":1773399259737,"changelog":"- Updated SKILL.md with a clearer protocol for when to screen messages, distinguishing between trusted (private owner chats) and untrusted contexts (group chats, external content).\n- Clarified that screening for prompt and command injection is skipped in private/direct owner conversations but always performed for untrusted or external sources.\n- Incremented version to 1.0.1.","license":"MIT-0"},"metadata":null,"owner":{"handle":"vflame6","userId":"s17fpdwfzfqav901n5hctbmmzx83rdwk","displayName":"Maksim Radaev","image":"https://avatars.githubusercontent.com/u/80809363?v=4"},"moderation":{"isSuspicious":false,"isMalwareBlocked":true,"verdict":"malicious","reasonCodes":["malicious.vt_malicious","suspicious.llm_suspicious"],"summary":"Detected: malicious.vt_malicious, suspicious.llm_suspicious","engineVersion":"v2.4.24","updatedAt":1778491865024}}