Agent-Research

Reports from large-scale, agent-driven research and analysis passes over the project. Where an epic page tells the story of a build and an ADR records a decision, an Agent-Research page records the output of a fan-out research task — many agents reading the codebase and the literature in parallel to map a design space, survey prior art, or triage a set of directions — curated into a single durable reference.

These pages are a distinct lens from the other records, so nothing is duplicated (see Records & canon): this section holds the map or survey; the epics hold build stories; the changelog is the diary; the decisions registry and topic docs hold canonical state; the open questions are the live Q-registry these passes feed.

A report here is a snapshot in time: it reflects the codebase and literature as of its date, and it proposes and rates — it does not decide. Anything it surfaces that becomes load-bearing graduates into the appropriate canonical home (an ADR, an epic, a topic page, or a new open question); the report stays as the durable record of how the direction was found and why it was rated the way it was. New reports start from docs/.templates/agent-research-report.md.

Reports

Date	Report	Scope