run - Agent runner for dispatching harness agents
Invokes an agent from a harness configuration, manages sprint directories, and handles inter-agent communication.
Tags
Updated: 2026-09-17Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Read harness configuration file
- Determine current sprint directory
- Gather input handoff files
- Determine expected output files
- Spawn agent via Agent tool
- Verify generated output files
- Update progress tracking file
- Evaluate feedback loop thresholds
- Suggest next workflow step
Inputs
- Agent name argument
- Harness configuration file
- Agent prompt files
- Sprint input handoff files
- Progress tracking file
Outputs
- Sprint output handoff files
- Updated progress tracking file
- Updated current sprint number
- Status and next step suggestions
Requirements
- Agent tool support
- Existing harness configuration file