Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
INTAKE now immediately triggers DISPATCHER in the same turn. Tasks start executing the moment they are queued. Heartbeat/cron is backup only.
What This Skill Does
A daemon-style task queue that persists across agent sessions. Users add tasks in natural language; a dispatcher executes them via subagents and polls on a heartbeat for retries and completions. The queue accumulates work over time and never auto-closes.
Combines persistent state, automatic retry logic, and concurrent subagent dispatch in a single queue file, so tasks survive session restarts without separate orchestration tooling.
When to use it
- Queuing multiple research tasks to run while you work on something else
- Batching file-processing jobs to execute in the background
- Delegating several unrelated tasks at once and checking results later
- Automatically retrying failed subagent tasks without manual intervention
- Tracking a persistent backlog of agent work across multiple sessions
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Do these for me: summarize last week's commits, check for broken links in the docs, and draft a release announcement
- 1Parses message into 3 distinct tasks
- 2Assigns sequential IDs (T-01, T-02, T-03) and appends them to task-queue.json
- 3Confirms added tasks to user and triggers dispatcher immediately in the same turn
- 4Dispatcher spawns up to maxConcurrent subagents to execute pending tasks in parallel
- 5Heartbeat polls for completion and notifies user as each task finishes or gets blocked
Three tasks run concurrently; user receives a done or blocked notification for each as subagents complete