Proposal: add an agent layer for artifact creation #593
Description
when you have a long exploratory conversation before formalizing a change, the full history gets passed to the command including abandoned ideas and back-and-forth. This bloats the context and can confuse the output.
After exploring with the user, Claude summarizes the relevant context and spawns an artifact-writer agent.
The agent uses existing commands (/opsx:continue, etc.) to create the artifact - but with clean, distilled context instead of the full
conversation history.
Commands stay as-is. The agent is just a context-filtering layer that delegates to them.