Nova Memory

本地存储和语义检索工作空间记忆，支持实体管理、标签搜索、图谱可视化及自动反思，无外部依赖，安全可靠。

This skill mostly implements a local memory engine and does not exhibit network exfiltration, but there are mismatches you should address before installing: (1) The code requires scikit-learn (sklearn) though the metadata claims no external deps — install scikit-learn in an isolated virtualenv/conda environment before use. (2) memory_olap.py's docs mention DuckDB but it actually uses sqlite3 — verify behavior if you expect DuckDB. (3) The import error message suggests using '--break-system-packages' to install scikit-learn; avoid that flag in shared/system Python environments — prefer a virtualenv. (4) The skill writes JSON files under /workspace/memory/nova-memory/ (and auto_reflect can write to MEMORY.md); ensure you are comfortable with data being stored there and run the skill in a sandboxed workspace if the data is sensitive. (5) The provided source appears mostly benign, but part of memory_core.py was truncated in the bundle — inspect the full save/load implementation before trusting it with sensitive data. If you intend to use this skill, test it in an isolated environment and confirm scikit-learn installation and file I/O behavior; if you need stronger assurance, request the publisher/homepage or a signed release.

Type: OpenClaw Skill Name: nova-memory Version: 1.0.0 The nova-memory skill bundle provides a local semantic memory and entity management system using TF-IDF and SQLite for analysis. The implementation in memory_core.py and memory_olap.py is purely local, restricted to the /workspace/ directory, and lacks any network capabilities, obfuscation, or dangerous execution sinks like eval() or system calls. The code logic is transparent and aligns exactly with the functionality described in SKILL.md.