Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
v3.5.3: 超寬鬆模式 - 強制記錄技術操作（安裝/配置/開發） Changes: - 降低長度閾值：30字 → 20字 - 降低重要性門檻：≥2 → ≥1 - 提升默認優先級：Normal → Important - 新增強制記錄關鍵詞：install, config, 開發, git, skill 等 - 確保所有技術相關對話都被記錄
What This Skill Does
Long-term memory framework for AI agents that stores, searches, and injects relevant memories into prompts automatically. Built on local JSON storage with keyword and semantic search, it supports priority tagging, time-based memory decay, and CJK text segmentation including Cantonese. Integrates with OpenClaw via a plugin hook that runs before each response.
Unlike RAG setups that require external vector databases or API calls, this system runs entirely on local files with no external dependencies.
When to use it
- Persisting user preferences across sessions
- Recalling project-specific context without re-explaining
- Tagging critical vs. normal information for selective retrieval
- Running memory search from external scripts via CLI JSON output
- Keeping agent context lean by decaying old or redundant memories
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: What display settings do I prefer?
- 1OpenClaw plugin fires before_prompt_build hook
- 2Plugin extracts query from event.prompt
- 3Searches local memory store for top 5 relevant entries
- 4Injects matched memories into prompt via prependContext
- 5Agent responds with personalized context already loaded
Agent answers referencing stored preference: dark mode, without user re-stating it