Mempalace Memory

基于MemPalace融合SuperMem增强层，提供自动hook注入、多样性重排、相似度去重和高效上下文记忆管理。

This skill is plausibly a local-memory enhancement, but it makes strong assumptions about the environment (hardcoded /Users/mars paths, Python site-packages, local Ollama at http://localhost:11434, and local ChromaDB paths). Before installing or activating: 1) Review the three Python scripts line-by-line (they will execute on your machine and access local files and DBs). 2) Replace or parameterize hardcoded paths (MEMPALACE_CLI, sys.path inserts) to match your system or remove them. 3) Ensure Ollama/chroma services are intended and secured (they are contacted locally via curl/subprocess). 4) Back up any local ~/.mempalace or ~/.super-mem data you care about; the scripts include deletion/forget functionality. 5) Prefer running first in a sandbox or test account, and only register the hook under ~/.openclaw/hooks when you trust the code. If you want, provide the full remainder of the truncated mempalace_cli.py (credential filter section) so I can re-evaluate any missing behaviors.

Type: OpenClaw Skill Name: mempalace-memory Version: 1.1.0 The skill bundle implements a sophisticated memory retrieval and storage system (MemPalace/SuperMem) for an AI agent. It features proactive security measures, including regex-based credential filtering in `mempalace_cli.py` and `super_mem_cli.py` to prevent the accidental storage of GitHub tokens or passwords. While the scripts utilize `subprocess` to interact with local binaries and `curl` to communicate with a local Ollama instance (localhost:11434), these actions are strictly aligned with the stated purpose of managing a local vector database and generating embeddings. The presence of hardcoded paths (e.g., `/Users/mars/`) indicates a personalized configuration rather than malicious intent.