compound-skill-orchestrator

The orchestrator coordinates multi-stage execution by invoking skills in sequence and passing outputs between stages, ensuring a unified final deliverable while keeping the main context clean through context isolation.

Context isolation uses a forked context for each pipeline stage to prevent context leakage. It ensures that multi-stage execution processes do not interfere with each other, keeping the main context clean.

Yes, you can integrate human-in-the-loop flexibility between automated pipeline stages. The orchestrator supports review and adjustments between steps, allowing manual intervention before the next skill executes.

Skill pipelines pass outputs by requiring each stage to write its output to a known file path. The orchestrator invokes skills in sequence, using these file outputs as inputs for the next stage.

Yes, you need existing skills to build an automated workflow. The orchestrator chains existing skills into a unified multi-stage workflow under a single command, rather than generating new skill logic.

Avoid using a multi-stage skill orchestrator when stages cannot write outputs to known file paths or do not require sequential dependency. It is designed specifically for end-to-end processes where one stage's output becomes the next stage's input.