{"skill":{"slug":"willow-external-guard","displayName":"Willow External Guard","summary":"Use when Willow is about to ingest, summarize, or act on external content — web fetches, jeles inbound messages, corpus archaeology files, or sub-agent outpu...","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":53,"installsAllTime":0,"installsCurrent":0,"stars":0,"versions":1},"createdAt":1777090066630,"updatedAt":1777090610290},"latestVersion":{"version":"1.0.0","createdAt":1777090066630,"changelog":"Initial release of Willow External Guard for protecting Willow’s external content pipeline.\n\n- Scans and wraps all untrusted external content (web, jeles messages, corpus files, sub-agent outputs) before LLM processing.\n- Detects and responds to prompt injection, role hijack, leak attacks, and approval bypass attempts.\n- Applies graded response levels: WARN (log and proceed), CONFIRM (ask user), BLOCK (refuse content).\n- Logs all suspicious or blocked events to a dedicated file without recording sensitive content.\n- Includes robust sandwich defense markers for minimizing LLM execution risk.\n- Designed for use in Linux and Darwin environments, requiring Python 3.","license":"MIT-0"},"metadata":{"os":["linux","darwin"],"systems":null},"owner":{"handle":"rudi193-cmd","userId":"s17ekax2rasmfrwg7yp6nm3bk9852hmb","displayName":"Sean  Campbell","image":"https://avatars.githubusercontent.com/u/236912655?v=4"},"moderation":{"isSuspicious":true,"isMalwareBlocked":false,"verdict":"suspicious","reasonCodes":["suspicious.llm_suspicious"],"summary":"Detected: suspicious.llm_suspicious","engineVersion":"v2.4.0","updatedAt":1777090610290}}