continuity

What This Skill Does

The Continuity framework transforms OpenClaw from a reactive tool into a proactive, evolving partner. Rather than treating every interaction as a blank slate, the Continuity skill enables asynchronous reflection. Once a conversation reaches an idle threshold (typically 30 minutes), the agent initiates a background reflection process. It parses the session to extract structured facts, user preferences, and evolving principles. This data is distilled into a persistent memory store, and the system intelligently generates follow-up questions to bridge the gap between sessions. By surfacing these questions at the start of a new interaction, Continuity ensures that the AI demonstrates genuine curiosity and continuity of thought, making the assistant feel more human and context-aware.

Installation

To begin, ensure you have the OpenClaw environment active. Install the skill directly from the repository using the following command in your terminal:

clawhub install openclaw/skills/skills/riley-coyote/memory-continuity

Once installed, verify the configuration by checking your environment variables. You can tune the CONTINUITY_IDLE_THRESHOLD to suit your specific workflow preferences, ensuring reflections trigger at a cadence that matches your usage patterns.

Use Cases

• Long-term Project Management: If you pause a coding session, Continuity will record your architectural decisions and unresolved bugs, prompting you with specific next steps when you return.
• Learning & Coaching: When discussing complex topics, the agent tracks your progress, remembering which concepts you found difficult and offering targeted review sessions.
• Relationship Building: For personal assistant roles, the agent stores preferences regarding communication style, tone, and specific personal constraints, ensuring it doesn't repeat unnecessary questions.

Example Prompts

• "Reflect on our last three sessions and tell me what the biggest bottleneck in my project is right now."
• "What questions are currently pending in my memory store that we should prioritize today?"
• "Show me the current status of my learned principles and relationship dynamics to see if anything needs updating."

Tips & Limitations

• Confidence Management: Always monitor the confidence scores. If the system logs a 'Speculative' memory, verify the information to prevent the AI from adopting incorrect assumptions.
• Environment Hygiene: Periodically review your questions.md file. While the agent handles pruning, manual oversight ensures the most meaningful questions rise to the top.
• Limits: The current implementation is optimized for text-based interactions; avoid overloading sessions with heavy multimedia data if you want the reflection phase to remain fast and accurate.