smart-agent-memory

What This Skill Does

Smart Agent Memory is a sophisticated, long-term memory system designed specifically for the OpenClaw AI agent ecosystem. It functions as a hierarchical context management engine that prevents token bloat by ensuring agents only access the specific information required for the current task. By utilizing a three-tiered storage approach—human-readable Markdown, structured JSON, and high-performance SQLite/FTS5—it provides a robust framework for state persistence. The system excels at 'Skill-based learning,' where successful tool executions or specific API quirks are archived as 'experience' memories, ensuring the agent becomes more efficient over time. It features a built-in session lifecycle management tool, allowing for clean session-start context loading and session-end summarization.

Installation

To install this skill, run the following command in your terminal: clawhub install openclaw/skills/skills/beyound87/smart-agent-memory

Once installed, you must perform a one-time initialization to configure your workspaces: node ~/.openclaw/skills/smart-agent-memory/scripts/memory-cli.js setup This command identifies your active workspaces and injects the necessary bootstrap configurations, enabling seamless cross-agent memory sharing.

Use Cases

• Project Continuity: Maintain complex project state and team preferences across different sessions and days.
• Skill Optimization: Store specific API parameter requirements or 'gotchas' for tools to prevent recurring errors.
• Contextual Awareness: Load relevant entity information or past discussions based on tags or timeframes, drastically reducing token consumption.
• Automated Reflective Learning: Use the 'learn' command to archive project outcomes, ensuring the agent avoids historical pitfalls in future iterations.

Example Prompts

• "@smart-agent-memory start session, tell me what we decided about the API authentication in our last meeting."
• "@smart-agent-memory remember that the Stripe API requires ISO-8601 formatted timestamps, tag this as stripe-integration."
• "@smart-agent-memory load context for project-x and give me a summary of lessons learned from the last 7 days."

Tips & Limitations

• Efficiency: Always call node $CLI index before loading deep context. This small index (<500 tokens) provides a roadmap, allowing you to fetch only the essential segments, saving significantly on costs and token limits.
• Persistence: Avoid storing massive binary blobs. The system is optimized for text-based context. If you need to store logs, use the built-in gc (Garbage Collection) functionality to archive old data and keep your memory footprint lean.
• Atomicity: Don't wait until the end of a long task to save information. Use remember or learn throughout the workflow. If an agent crashes, you won't lose the progress made in the middle of the task.
• Structure: Leverage tags consistently. The effectiveness of the system relies on your ability to categorize memories logically so they can be retrieved instantly via the context command.