subagents

Parallel batch processing for independent file refactoring is handled by launching separate subagent processes with explicit instructions for each non-dependent unit. You issue pi commands with the same model but different instructions per task to achieve concurrent execution.

You can run parallel analyses across datasets by splitting work into non-dependent units and launching separate subagent processes for each. This orchestration approach allows independent analyses to execute concurrently rather than sequentially.

The maximum supported for batch file comparisons is ten concurrent subagent processes. Each process runs independently with its own explicit instructions, and sessions are saved automatically upon completion.

Orchestrating independent subagents to compare files in parallel requires issuing pi -p --model commands with explicit, distinct instructions for each subagent. Each process handles a separate non-dependent file comparison unit automatically.

No external dependencies are required to launch parallel subagents for independent tasks. The orchestration relies solely on issuing pi commands with explicit model specifications and separate instructions for each concurrent process.

You should avoid parallel subagent orchestration when tasks are dependent on each other, as this approach is designed exclusively for non-dependent units. If your workflow requires sequential execution or shared state between steps, parallel processing will not work correctly.