Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Version 2.0.0 is a major redesign of the interaction decision and output format: - Replaces old yes/no and debug output with a single decision field; on yes, provides both an interaction type and description for more natural, guided outreach. - Refines interaction types to emphasize casual, human-like chat (e.g., playful openers, context-aware follow-ups, light shared observations). - Shifts to concise, easy-to-ignore, non-transactional message guidance avoiding "just checking in" and similar phrasing. - Requires real context for data-aware interactions; if unavailable, defaults to more general, low-pressure messages or retries for a non-data type. - Omits probability and roll details from the output for cleaner integration and simpler error handling. - Clarifies contract rules, error handling, and offers more examples of acceptable outputs.
What This Skill Does
Runs a Python script to decide whether an agent should send a spontaneous casual message during a scheduled check-in. On a yes decision, it returns an interaction type and short guidance for drafting the message. Supports data-aware interactions when live context like weather, calendar, or smart-home state is accessible.
Keeps proactive agent messages from feeling patterned by randomizing both timing and interaction style rather than using a fixed schedule or template.
When to use it
- Scheduling periodic friendly pings from a personal assistant agent
- Varying proactive outreach so it feels less mechanical
- Injecting calendar-aware nudges before upcoming events
- Sharing a weather-relevant comment on relevant mornings
- Pulling smart-home status into a casual one-liner
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: agent runs its scheduled check and calls the interaction designer
- 1Run python3 scripts/random_interaction_designer.py
- 2Read decision field from JSON output
- 3Stop if decision is no
- 4If decision is yes, read interaction_type and interaction_description
- 5Fetch relevant live context if the interaction type is data-aware, then draft a brief casual message following the description
A short one or two line chat message sent to the user, such as a playful opener or a calendar nudge, with no mention of the scheduling or random process behind it