Memory Master

AI Memory System with Iterative Compression & Lineage Tracking - Hermes Agent Inspired

This repo looks like a plausible local memory/compression library, but several red flags merit caution before installing or running it: 1) Backup first: backup ~/.openclaw/workspace/, AGENTS.md, MEMORY.md and any agent config and your memory.db file before running any scripts or schema updates. 2) Inspect scripts: open scripts/init.js and scripts/detect.js and any code that performs fs.write/File/rename, child_process.exec, network calls (fetch, http(s), axios), or sqlite/DB writes. Look for explicit writes to AGENTS.md, ~/.agents, or other agent config paths. 3) Verify automatic migrations: the README/README_v4.2.0 states the package can automatically merge rules into AGENTS.md and auto-write files — if you do not want third-party code changing agent-wide rules, do not run migration/init scripts. 4) Run in a sandbox: if you want to test, run the code in an isolated environment (VM/container) and not on your primary agent host, until you confirm no unwanted behavior. 5) Check network/LLM usage: search the code for calls that contact external LLM APIs or remote endpoints. If it needs an API key, prefer explicit configuration rather than hardcoded endpoints. If the code performs 'auto-learning' from the web, be aware it may make outbound requests. 6) Confirm install path: the skill metadata said 'instruction-only' but package.json includes an init script; clarify whether the platform will execute included scripts automatically. If the platform will run init scripts, they should be audited first. 7) If you want this skill but not its migrations: manually copy only the compressor/manager code you trust into your project and avoid running the provided init/migration scripts. If you want, I can (a) summarize the specific file-level places to inspect (search for fs.writeFile/rename/exec/fetch/axios/sqlite3 usage), or (b) scan the included files for those patterns and point to exact lines that perform filesystem or network operations.

Type: OpenClaw Skill Name: openclaw-memory-master-v4 Version: 4.2.0 The openclaw-memory-master skill bundle is a comprehensive and well-documented memory management system for AI agents. It implements a multi-layered storage architecture (L0/L1/L2) in `src/layers/`, automated memory classification in `src/memory-classifier.js`, and an iterative compression system in `src/compressors/aaak-iterative-compressor.ts` to maintain long-term context. The scripts (`scripts/init.js` and `scripts/detect.js`) handle workspace setup, backups, and resource monitoring locally. The instructions provided to the agent via `SKILL.md` and templates are entirely consistent with the stated goal of enhancing an agent's memory and do not exhibit any signs of malicious intent, unauthorized data exfiltration, or harmful prompt injection.