Glossary
AI agent product memory terms
Plain-language definitions for how RAAV coordinates coding agents, founders, and durable product truth.
Product memory
The durable product truth a team and its coding agents share — ICP, problem, promise, scope, requirements, risks, and decisions.
Memory candidates
Inferred product facts agents propose after work — kept out of confirmed memory until a founder promotes or rejects them.
Founder review queue
The control surface where founders approve, defer, or reject agent proposals, memory updates, and decisions.
Context envelope
The structured task context RAAV attaches when a founder messages an agent — task, branch, claim, acceptance criteria, review policy, and submit command.
Branch lanes
Parallel work tracks that tie tasks, git branches, and agent ownership together so multiple agents do not collide on the same surface area.
Agent claims
Explicit file or scope ownership an agent takes before editing, recorded in the RAAV ledger for audit and conflict detection.