continuity

What This Skill Does

The Continuity framework transforms OpenClaw from a reactive chatbot into an evolving partner. By automating the reflection process, it bridges the gap between disconnected conversation sessions. The skill functions by monitoring user activity; once an idle threshold is met, it triggers a reflection sequence that synthesizes the conversation history into structured, categorized memory types. It identifies key facts, evolving preferences, and latent commitments, then generates genuine, curious follow-up questions. When you return, it surfaces these insights, ensuring the AI maintains context, learns from past successes, and refines its behavior based on your specific needs and communication style.

Installation

To integrate Continuity, run the following command within your environment: clawhub install openclaw/skills/skills/riley-coyote/continuity-framework

After installation, configure your HEARTBEAT.md to trigger the reflection process after 30 minutes of idle time. Ensure that your memory/ directory structure is accessible, as the skill creates and maintains files like identity.md and questions.md to manage your long-term relationship context.

Use Cases

Continuity is ideal for complex, long-term project management where context is often lost between days. It excels for users who want their AI to learn personal preferences over time without repeating themselves. It is also highly effective for collaborative brainstorming, as it remembers the trajectory of ideas and can spontaneously reintroduce previously tabled topics that have gained new relevance.

Example Prompts

1. "Reflect on our work yesterday and identify any loose ends or questions we still need to resolve for the project architecture."
2. "Show me the current state of our memory, specifically any preferences you have inferred about my coding style."
3. "Start our session today by summarizing what we learned during the last session and surfacing any priority questions for our next steps."

Tips & Limitations

To maximize the utility of Continuity, ensure you provide enough interaction volume (adjust the CONTINUITY_MIN_MESSAGES setting) to avoid overfitting on noisy data. Be aware that the confidence scores are essential; review 'speculative' memories periodically to ensure the agent hasn't misinterpreted a specific comment as a long-term preference. The skill is designed to improve with more diverse interactions, so treat the reflection logs as a tool for iterative refinement of your AI agent's personality and competence.