Memory Ops

What This Skill Does

Memory Ops, developed by ianleme, is the backbone of operational persistence for the OpenClaw agent ecosystem. It serves as the primary interface between the agent and a PostgreSQL database leveraging the pgvector extension. By enabling semantic search and persistent storage, Memory Ops ensures that OpenClaw does not suffer from short-term context loss. It enforces a strict protocol where every interaction cycle involves querying historical context, capturing user intent, recording delegation handoffs to specialized sub-agents (like Alfred or Prompt Improver), and logging detailed audit trails for every read/write operation.

Installation

To integrate this skill into your environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/ianleme/memory-ops Ensure your local or remote PostgreSQL instance is configured with the vector extension and that the necessary schema files provided in the source repository are initialized to support the memories and memory_audit tables.

Use Cases

• Long-term User Profiling: Storing user preferences, technical constraints, and evolving project goals to personalize future interactions.
• Agentic Handoffs: Managing complex workflows where tasks are passed between OpenClaw and specialized agents, ensuring all parties have the required context to proceed without redundant information gathering.
• Operational Auditing: Maintaining a transparent log of what the AI read, what it wrote, and why, essential for debugging agent behavior in production environments.
• Knowledge Retrieval: Using semantic search to fetch relevant historical facts or previous decisions when the user asks "What did we decide about the architecture last week?"

Example Prompts

1. "Check the memory for our project requirements and summarize the current status of the database migration."
2. "I am delegating the task of prompt optimization to the Prompt Improver; please log this handoff in the memory audit."
3. "Save my preference for using PostgreSQL over MongoDB for this project into the memory database."

Tips & Limitations

• Dimension Alignment: If you switch embedding models, ensure the vector(1536) column definition in memory_ops_template.sql is updated to match your model's output dimensionality.
• Privacy First: While the tool is powerful, avoid writing plain-text secrets like API keys or passwords directly into the memories table. Use external secret management systems for sensitive credentials.
• Audit Compliance: Always verify that your memory_audit table is properly indexing the timestamp column to maintain query performance as the history grows over time.