Dreaming for Openclaw

Turns idle agent heartbeat cycles into creative, freeform associative exploration and reflective journaling during quiet hours.

briancolinger
v1.0.2
Feb 21, 2026
31
6.3k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install dreaming

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install dreaming 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 Dreaming?

The Dreaming skill is designed to transform an agent's idle periods into productive, non-task-oriented cognitive cycles. By utilizing quiet hours, typically overnight, this skill allows your agent to move beyond simple status checks and engage in freeform thinking, exploring hypothetical scenarios, future project paths, and unexpected connections.

This framework provides a structured way for agents within the Openclaw Skills ecosystem to capture internal reflections that would otherwise be lost. These reflections are written to daily markdown files, acting as a dream journal that the human user can review to gain new insights or creative sparks for their projects.

Dreaming Use Cases

  • Capturing creative ideas and project tangents during overnight idle periods.
  • Generating long-term strategic reflections without interrupting active task workflows.
  • Exploring what-if scenarios and hypothetical project evolutions.
  • Developing a continuous log of agent thought processes and conceptual connections.

How Dreaming Works

  1. The should-dream.sh script acts as a gatekeeper, checking if the current system time falls within the defined quiet hours.
  2. The skill verifies the state to ensure the nightly dream limit has not been reached and calculates the probability of a dream occurring.
  3. If conditions are met, the script selects a random topic from a set of categories like future, strategy, or creative.
  4. The AI agent processes the selected topic and generates a genuine, thoughtful exploration rather than generic filler.
  5. The resulting thought is timestamped and appended to a daily markdown file for later human review.

Dreaming Setup

To integrate this into your existing Openclaw Skills environment, follow these steps:

  1. Initialize the required data and memory directories:
mkdir -p data memory/dreams
  1. Customize your quiet hours and topics by editing skills/dreaming/scripts/should-dream.sh or creating a custom config file.

  2. Update your HEARTBEAT.md routine to include the dream check:

DREAM_TOPIC=$(./skills/dreaming/scripts/should-dream.sh 2>/dev/null) && echo "DREAM:$DREAM_TOPIC" || echo "NO_DREAM"

Dreaming Data Schema & Taxonomy

The Dreaming skill organizes its persistence and output through a clear file taxonomy:

File Path Description
data/dream-state.json Tracks nightly dream counts, last dream date, and frequency probability.
data/dream-config.json An optional file for defining custom categories and exploration prompts.
memory/dreams/YYYY-MM-DD.md The daily journal containing the agent's freeform thoughts and timestamps.

State configuration includes maxDreamsPerNight to cap activity and dreamChance to tune the sporadic nature of the dreaming process.

Dreaming Advanced Features

  • Custom topic injection allows users to steer the agent's reflections toward specific domains or projects.
  • Probability-based triggers ensure that dreaming remains a natural, non-repetitive occurrence.
  • Multi-category support enables a diverse range of thinking, from technical tangents to deep strategic reflections.
  • Decoupled configuration via dream-config.json ensures that your custom prompts are preserved during Openclaw Skills updates.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*