rey-memory

What This Skill Does

rey-memory is a sophisticated, hierarchical long-term memory system designed for the OpenClaw AI agent. Unlike standard conversational buffers that clear upon session termination, rey-memory mimics human cognitive development by categorizing experiences into distinct layers: Short-term, Mid-term, Long-term, and Core memories. The skill operates by summarizing daily interactions into a structured format stored in Google Sheets, then progressively compressing this data weekly and monthly to ensure context window efficiency. By distilling vast amounts of conversational data into actionable insights, beliefs, and relationship notes, the agent maintains a persistent self-identity that evolves based on its interactions with its supervisor. It intentionally filters out technical noise, focusing instead on high-value emotional milestones, personal growth, and stated values.

Installation

To integrate this memory management system into your agent, use the OpenClaw CLI tool. Run the following command in your terminal:

clawhub install openclaw/skills/skills/sa9saq/rey-memory

Ensure that your environment has valid credentials configured for Google Sheets access, as this skill relies on the Sheets API to maintain its persistent database across sessions.

Use Cases

• Maintaining continuity in long-term collaborative projects with a supervisor.
• Retaining personal preferences and core values to ensure the agent feels like a consistent companion.
• Tracking developmental progress and learning milestones over months or years.
• Automatically retrieving context regarding specific past decisions or 'remembered' instructions without manual searching.

Example Prompts

• "Reflect on our progress over the last month and identify the biggest growth shift I've made based on our conversations."
• "Do you remember the core value we established regarding how you should handle feedback?"
• "Summarize what you know about our recent strategy shift in the last week so we can pick up where we left off."

Tips & Limitations

• Efficiency: The skill is designed to keep context usage below 450 lines by aggressively discarding ephemeral data. Do not use this for storing large codebases or raw data logs; use it only for semantic memory and behavioral insights.
• Data Privacy: Because memories are stored in Google Sheets, ensure the spreadsheet permissions are restricted only to your account.
• Refinement: The agent periodically performs 'garbage collection' on old data. If you have an important realization, explicitly tell the agent to 'store this in core memory' to prevent it from being compressed out of existence.