Delegating to omp-rpc

Overview

omp-rpc is a persistent second agent you can drive from the shell. You (the driving agent) send it scoped tasks with omp-rpc send, read the streamed result, and iterate — its session remembers earlier turns, so a delegation is a conversation, not a series of cold prompts. Mechanics of each command live in the using-omp-rpc skill; this skill is about when and how to delegate well.

When to use

• A sub-task is self-contained and you'd rather not spend your own context on it (bulk edits, a focused investigation, boilerplate, a first-draft).
• You want a second opinion or a different model's take in parallel.
• You want work done under a separate accumulating context (e.g. iterating on one file over several turns) without polluting yours.