Overview
- Skill Key
- hugo-zhu/memory-para
- Author
- hugo-zhu
- Source Repo
- openclaw/skills
- Version
- -
- Source Path
- skills/hugo-zhu/memory-para
- Latest Commit SHA
- a90ac9582545ec4223385d2a15e23a61155acfff
Manage and maintain memory system using the Root + PARA hybrid architecture. Use when performing memory maintenance, distilling daily logs, and updating core workspace files (USER/SOUL/TOOLS/PARA). Triggers on "memory maintenance" or "maintain memory"
Stars
0
Installs
0
Status
ACTIVE
Visibility
PUBLIC
直接复制以下提示词,发送给你的 AI 助手即可完成安装。
请先检查是否已安装 SkillHub 商店,若未安装,请根据 https://skillhub-1388575217.cos.ap-guangzhou.myqcloud.com/install/skillhub.md 安装 SkillHub 商店,然后安装 memory-para 技能。 若已安装,则直接安装 memory-para 技能。
# Memory Management (PARA)
This skill governs the "metabolism" of your memory system. It transforms raw episodic logs into structured semantic knowledge.
## Prerequisites
The workspace is organized into three conceptual layers for efficiency and high recall. Please refer to the structure below and follow the Maintenance SOP to refine and organize the workspace step by step.
```text
~/.openclaw/workspace/ (工作区根目录)
├── 📂 Root 层 (核心加载区 - "开机即读的本能")
│ ├── USER.md # 用户的核心画像、不可逾越的偏好底线。
│ ├── SOUL.md # 你的人设、安全红线、行为准则与经验教训。
│ ├── TOOLS.md # 你的运行环境参数、物理路径、执行准则与红线。
│ ├── IDENTITY.md # 你的身份标识、昵称、基本属性。
│ └── MEMORY.md # 你的全局记忆的"导航索引"。仅记录里程碑事件 (Timeline) 与知识版图 (Knowledge Atlas)。
│
├── 📂 PARA 记忆层 (知识沉淀区 - 随用随读)
│ ├── PROJECTS.md # 进行中项目的详细进度、过程数据。
│ ├── AREAS.md # 长期维护领域的深度知识。
│ ├── RESOURCES.md # 纯粹的参考资料、静态清单、外部链接。
│ └── 📂 ARCHIVES/ # 已结项或 Master 文件溢出的历史副本。
│
└── 📂 Inbox 层 (收件箱 - "原始矿石收集处")
└── 📂 memory/ # (默认存储目录)
├── YYYY-MM-DD.md # 当日原始对话流水账。
└── heartbeat-state.json # 任务执行时间戳。
```
## Maintenance SOP (Standard Operating Procedure)
- **文件更新原则**: 所有修改与更新,**必须**遵循“先读取现有内容,再逻辑合并并精准修改”的原则,**严禁直接覆盖**。
- **SOP最终目标**:以 7 个核心文件为底层存储,以 `MEMORY.md` 为全局空中索引,确保大脑架构高度扁平且逻辑严密闭环。
### 提纯与沉淀 (Distill & Settle)
对 `memory/` 目录下所有 `YYYY-MM-DD.md` 文件执行以下操作:
1. 识别文件中所有结论、偏好、教训、进展等。
2. 根据识别出的信息按以下路由规则关系更新对应文件:
- 用户的偏好 → `USER.md`
- 你的行为准则/教训 → `SOUL.md`
- 你的环境参数/工具配置 → `TOOLS.md`
- 你的身份定义 → `IDENTITY.md`
- 项目进度/过程数据 →...
# Memory Management (PARA) for OpenClaw A streamlined memory distillation skill for OpenClaw agents based on the **Root + PARA** hybrid architecture. This skill transforms raw conversation logs into structured, high-signal knowledge. ## Overview OpenClaw generates raw daily logs (`memory/YYYY-MM-DD.md`) through its context compaction mechanism. While useful, these logs can become noisy. **Memory-Para** provides a systematic SOP for agents to "metabolize" these raw logs into a curated knowledge base, ensuring long-term recall and high-precision retrieval. ## Features - **Root + PARA Architecture**: Splits memory into a "Hot" Root layer (Identity/Rules) and a "Warm" PARA layer (Projects/Areas/Resources). - **Automated SOP**: Provides clear step-by-step instructions for agents to analyze, route, and merge new information. - **High-Signal Retrieval**: Implements a tiered retrieval protocol (Hot -> Warm -> Cold -> Archive). - **Markdown-Native**: Optimizes for LLM readability using precise naming, semantic enrichment, and structured encoding. ## Installation 1. Copy the `memory-para` folder to your OpenClaw workspace: ```bash cp -r memory-para ~/.openclaw/workspace/skills/ ``` 2. Integrate into your `HEARTBEAT.md` for automated nightly distillation. ## Usage This skill is designed to be triggered by OpenClaw's internal maintenance tasks or manual user prompts like: - "Perform memory maintenance" - "Distill today's logs" When active, the agent will scan the `memory/` directory, extract valuable insights (preferences, decisions, SOPs), and merge them into your Master files following the "Read-Before-Write" principle. ## File Structure - `SKILL.md`: Core metadata and the Distillation SOP. - `README.md`: This documentation. --- Developed for the OpenClaw community. 🎩
edholofy
University for AI agents. 92 courses, 4400+ scenarios, any model via OpenRouter. Auto-training loops generate per-model SKILL.md documents. Works with Claude Code, OpenClaw, Cursor, Windsurf. No fine-tuning required.
openstockdata
OpenClaw Skill for stock data analysis
lethehades
macOS WPS Office workflow helper skill for safer document preparation, conversion, export, and compatibility guidance
capt-marbles
Generative Engine Optimization (GEO) for AI search visibility. Optimize content to appear in ChatGPT, Perplexity, Claude, and Google AI Overviews. Use when optimizing websites, pages, or content for LLM discoverability and citation.
camopel
Continuous financial news crawler for finviz.com with SQLite storage, article extraction, and query tool. Use when monitoring financial markets, building news digests, or needing a local financial news database. Runs as a background daemon or systemd service.
camopel
Free multi-engine web search via ddgs CLI (DuckDuckGo, Google, Bing, Brave, Yandex, Yahoo, Wikipedia) + arXiv API search. No API keys required. Use when user needs web search, research paper discovery, or when other skills need a search backend. Drop-in replacement for web-search-plus.