Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Queues decision plans for async human review via Arbiter Zebu. Agents push a set of choices with context and options, then a human reviewer selects answers. Supports batching multiple decisions in one plan, priority levels, and notifications to unblock waiting agents.
Batches multiple related decisions into one reviewable plan rather than interrupting reviewers with separate one-off questions.
When to use it
- Reviewing auth strategy options before starting API work
- Getting database choice approval during sprint planning
- Batching architectural tradeoffs into a single review request
- Pausing agent work until a human approves a deployment approach
- Notifying a blocked agent when all decisions are complete
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Which auth strategy and database should we use for the NFT marketplace API?
- 1Calls arbiter_push with a plan containing two decisions: auth strategy (JWT, sessions, OAuth) and database (PostgreSQL, MongoDB)
- 2Sets priority to 'high' and notify to the blocked agent's session key
- 3Polls status with arbiter_status using the project tag until review is complete
- 4Retrieves final answers with arbiter_get
- 5Proceeds with implementation using the selected options
Plan completed: auth-strategy=jwt, database=postgresql. Agent resumes API implementation.
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.