What problem does it solve?
When working on tasks, you may have specialized AgentField agents installed on your machine that could handle the work, but discovering them, learning their input contracts, dispatching jobs, and monitoring runs requires knowing the right CLI commands and API endpoints. This Skill provides the complete operational playbook for offloading work to installed AgentField subharnesses instead of doing everything inline.
Core Features & Use Cases
- Runtime Discovery: Find which agents and reasoners are installed and running using
af ls -e, af agent search, and the discovery capabilities API, with health-status checks to avoid dispatching to dead nodes.
- Contract-First Dispatch: Fetch exact input schemas with
af call <node>.<reasoner> --schema before calling, dispatch asynchronously, and monitor runs via af wait, af tail, or batch-status polling.
- Failure Handling & Reporting: Diagnose missing API keys, wedged runs, and silent failures with a structured troubleshooting table, then report results with duration, cost windows, and live run URLs.
- Use Case: You need to review ten pull requests. The Skill checks the fleet, finds the pr-af agent, dispatches all ten reviews concurrently, polls them as a group, and reports each result with its run URL.
Quick Start
Ask the assistant to check which AgentField agents are installed and offload the current task to a covering subharness.