subagent - Run one-shot and parallel provider subagents
Dispatch headless provider LLM subagents for one-shot or flat parallel independent tasks and return their results.
Tags
Updated: 2026-10-02Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Run headless subagents
- Fan out independent tasks
- Select provider models
- Limit task duration
- Cap task spending
- Return JSON results
Inputs
- Task prompts
- Provider configuration
- Provider or model selection
- Prompt files
- Timeout settings
- Budget and turn limits
- Team lead session ID
Outputs
- Subagent result text
- JSON success or failure envelope
- Usage and cost metadata
- Provider and model metadata
- Session identifier
- Failure status
Requirements
- cc-fleet installed on the execution host
- Host-executing Bash access
- An enabled configured provider or Claude Code login
- Access to referenced shared documentation