Braindump

Memory that survives the session. One dense fact per line, appended to a topic file, grepped back later.

Manual triggers:

• "braindump this" / "save this to memory" — store
• "use your brain" / "remember anything about..." — retrieve

Otherwise use it proactively, per below.

Storage

.agents-workshop/braindump/<topic>.md — one file per topic, you choose them. Group by what a future session would search for, not by date or author.

Outside a scaffolded repo, use ~/.braindump/.

Committed, never ignored. Shared means shared with the next clone.

No index, no database, no CLI. Files and grep.