Memory System V2 1.0.0

Fast semantic memory system with JSON indexing, auto-consolidation, and <20ms search. Capture learnings, decisions, insights, events. Use when you need persi...

This skill appears to do what it claims: a local, file-based memory CLI built in bash that only needs jq. Before installing or using it, note: (1) it will create and write files under $HOME/clawd/memory (daily logs, index, consolidated summaries) — review or back up that directory if you care about its contents; (2) there are no network calls or credential uses, so it does not exfiltrate secrets by design; (3) implementation caveats: the script uses inlined JSON construction for index entries and could produce malformed JSON if you pass content with unescaped quotes/newlines (test with safe input first), and it uses both BSD and GNU date fallbacks which may behave differently on some systems — test on your OS; (4) if you plan to integrate this with autonomous agents, remember the skill does not include automatic agent hooks—adding agent-level auto-capture should be done intentionally and reviewed. If you want extra caution, review the memory-cli.sh source locally and run it in a disposable environment before adding it to production workflows.

Type: OpenClaw Skill Name: memory-system-v2-1-0-0 Version: 1.0.0 The OpenClaw AgentSkills bundle 'memory-system-v2' is classified as benign. The `memory-cli.sh` script and associated documentation (`SKILL.md`, `README.md`, `docs/*`) implement a local, file-based memory system. All operations are confined to the agent's `$HOME/clawd/memory` directory, involving standard file system operations and `jq` for JSON processing. There is no evidence of data exfiltration, malicious execution (e.g., `curl|bash`), persistence mechanisms, or obfuscation. The `SKILL.md` provides clear instructions for the agent to use the memory system, without any prompt injection attempts designed to subvert the agent's behavior, hide actions, or access unrelated sensitive data. Input sanitization for `jq` arguments is handled appropriately using `--arg` and `--argjson`, mitigating common injection risks.