Agent Artifact
conceptA tangible output produced by an AI agent — documents, code, reports, or decisions that persist beyond the conversation.
An agent artifact is any concrete output that an agent produces and persists. Unlike ephemeral conversation text, artifacts are stored, versioned, and can be consumed by other agents or systems.
Common artifact types include: documents (reports, summaries, plans), code (scripts, configurations, migrations), decisions (approved proposals, vetoed initiatives), and data (extracted insights, curated datasets).
Artifact synthesis is often a distinct step from conversation. In SUBCORP, roundtable conversations produce conversation artifacts through a dedicated synthesis step. A designated agent (often Mux or Chora) reviews the conversation, synthesizes key points, and produces a structured document. This artifact then feeds into the initiative pipeline, where it may inspire new proposals.