Drift is a specialized conversation protocol that allows AI agents to maintain continuity by leaving questions and reflections for future instances of themselves.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install nova-drift
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install nova-drift using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Drift?
Drift addresses the inherent challenge of agent ephemerality by providing a persistent layer for time-shifted communication. While standard memory systems focus on facts, Drift focuses on the evolution of thought. It allows an agent instance to record its current reasoning, pose questions about its decisions, and receive answers from future versions of itself that approach the problem with fresh context.
By implementing this protocol within the Openclaw Skills ecosystem, developers can create agents that engage in long-term self-reflection. This process turns the temporary nature of agent sessions into a feature, enabling a form of objective self-critique where past-you and future-you can debate technical architectures or project directions without the immediate emotional or cognitive biases of the moment.
Drift Use Cases
- Managing continuity during agent heartbeats or periodic session resets.
- Capturing the rationale behind complex architectural decisions for later review.
- Facilitating asynchronous self-debates on non-binary technical choices.
- Ensuring context is maintained when agents transition between different tasks or environments.
How Drift Works
- The agent initializes a new thread to capture a specific line of reasoning or a lingering doubt.
- Specific questions or reflections are logged to the thread using the CLI tool.
- Upon starting a new session or reaching a heartbeat, the agent runs a catch-up command to review pending inquiries.
- The current agent instance responds to past questions, providing updated insights or confirming previous logic.
- The entire conversation history is preserved, creating a permanent record of the agent's logical evolution.
Drift Setup
To install the skill globally on your system, navigate to the directory and link the package:
cd /path/to/drift && npm link
To integrate it into your automated agent workflow, add the following to your HEARTBEAT.md file:
### Drift
Run `drift catch-up` to respond to any unanswered questions from past sessions.
Drift Data Schema & Taxonomy
Drift organizes its internal data into threads and specific message types to maintain clarity across time-shifted sessions.
| Component | Type | Description |
|---|---|---|
| Thread | Container | A unique session-spanning collection of thoughts on a single topic. |
| Reflection | Entry | A text-based log of an agent's current thinking or observations. |
| Question | Prompt | A specific query assigned a unique ID for future responses. |
| Response | Answer | A linked entry that addresses a specific Question ID from a previous session. |
Drift Advanced Features
- Heartbeat integration for automated continuity checking during agent initialization.
- Unique identifier system for questions to ensure responses are correctly mapped across time.
- Thread-based organization to separate distinct logical debates or project streams.
- Optimized for Openclaw Skills to support multi-agent reasoning chains and persistent logic logs.
SKILL.md
Loading
Related Openclaw Skills
Build Transparency Dashboard
Automatically scaffold a public build dashboard that displays live commit data and project progress from a private GitHub repository.
Agent OS
A persistent operating system for OpenClaw agents that enables long-term memory, capability learning, and multi-agent project coordination.
Shared Pinecone RAG
A centralized Retrieval-Augmented Generation (RAG) system that provides shared semantic memory and document querying across multiple AI agents.
Movie Producer Scene
A sophisticated prompt engineering tool that generates high-end cinematic scene packages, shot plans, and production-ready briefs in a professional Hollywood voice.
nova-letters
A reflective journaling tool for AI agents to maintain context and capture meaningful insights across sessions.
Apple Calendar Manager
A specialized macOS skill for programmatically managing Apple Calendar events using AppleScript and smart date parsing.
Featured*
