Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Defines recurring maintenance routines for agent environments across daily, weekly, and monthly cadences. Covers secret scanning, dependency audits, memory compaction, prompt injection filtering, and system health checks. Includes a heartbeat dispatcher that runs local checks before escalating to cloud LLMs.
The heartbeat dispatcher runs health checks and email triage locally before escalating to a cloud LLM, cutting unnecessary API calls and token costs.
When to use it
- Running a secret scan before merging a pull request
- Rotating credentials after a suspected key exposure
- Filtering incoming emails through prompt injection guard
- Cleaning up stale sub-agent sessions at end of a work session
- Compacting a week of daily logs into long-term memory
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Run the heartbeat check
- 1Execute health-check.sh to capture system vitals (CPU, RAM, disk)
- 2Read heartbeat-state.json to identify overdue tasks
- 3Run secret-scan.sh if the 24h window has elapsed
- 4Triage AgentMail inbox through prompt-guard using the local Reef LLM
- 5Output HEARTBEAT_OK if nothing requires escalation, or emit an alerts JSON payload if action is needed
HEARTBEAT_OK with no alerts, or a JSON alerts payload listing overdue tasks or flagged content for human review