Spin up a background agent to do the research, so you keep working while it reads. When you are already a subagent, do the research yourself with the same job.
Its job:
- Investigate the question against primary sources (official documentation, source code, specifications, first-party APIs), not a secondary write-up of them. Follow every claim back to the source that owns it.
- Write the findings to a single Markdown file, citing each claim's source.
- Save it where the repository already keeps such notes, as
AGENTS.mdnames them; match the existing convention. If there is none, put it somewhere sensible, say where, and add one line to the nearestAGENTS.mdsaying what the notes hold and when to read them.