Spin up a background agent to do the research, so you keep working while it reads.

Its job:

Load the web tools before reading anything. WebSearch and WebFetch are deferred — only their names are advertised, so their schemas must be pulled in with ToolSearch (query select:WebSearch,WebFetch) before either can be called. Skip this and the agent finds no fetch tool, silently falls back to curl, and researches badly: raw HTML instead of markdown, no search step so sources are guessed at rather than found, outright failure on JS-rendered and bot-protected pages, and one shell spawned per fetch. Reserve curl for what it is actually good at — a plain JSON API, or checking an HTTP status.
Investigate the question against primary sources — official docs, source code, specs, 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 repo already keeps such notes; match the existing convention, and if there is none, put it somewhere sensible and say where.