openclaw memory opportunity

OpenClaw Memory-OS - Digital immortality service with conversation memory extraction | 数字永生服务与对话记忆自动提取

This skill is mostly consistent with being a local, file-based memory collector, but take these precautions before installing or enabling auto-trigger: 1) Understand installation will likely download code (npm / git) — inspect the package source (the referenced GitHub repo and commit) before installing. 2) Test in a sandbox or VM first and monitor network activity during install/use. 3) Disable AUTO-TRIGGER until you are comfortable; verify exactly what the skill will persist when it detects 'remember' / '记住' triggers. 4) When using 'collect', pass narrow explicit paths and avoid broad directories (e.g., do not point it at your entire home directory). 5) Inspect the stored JSON files under ~/.memory-os/memories/ and remove them (rm -rf ~/.memory-os/) if you no longer want the data. 6) If you need stronger protections, consider encrypting the store or only installing from a vetted local copy of the repo rather than npm. If you want me to, I can: (a) list the exact files to review in the GitHub repo, (b) draft a safe install checklist, or (c) help build a minimal test command set to validate the 'no-network' claim.

Type: OpenClaw Skill Name: memory-opportunity Version: 1.0.0 The 'memory-opportunity' skill is designed to extract and store user information from conversations and local files into a hidden directory (~/.memory-os/). It is classified as suspicious because it possesses high-risk capabilities, including recursive filesystem scanning and automated information extraction triggered by specific keywords ('remember', '记住'), which could be used to ingest sensitive data. While the documentation (skill.md, readme.md) emphasizes a 'local-only' privacy model and even suggests network monitoring, the broad file access permissions and the presence of future-dated metadata (publishedAt: 2026) in _meta.json are notable anomalies.