memory-maintain
memory-maintain
Periodic housekeeping for the ai-context/ memory layer: archive old changelog entries, separate resolved known-issues, regenerate the index, and detect CLAUDE.md configuration gaps.
Triggers: /memory-maintain, maintain memory, memory housekeeping, clean ai-context, ai-context cleanup
Purpose
The ai-context/ memory layer has no automatic cleanup mechanism. Over time, changelog-ai.md grows unbounded (expensive context window cost), known-issues.md accumulates resolved items without formal separation, and there is no single-page index explaining what files exist. memory-maintain handles this periodic backlog cleanup. It is complementary to memory-init (generates from scratch) and memory-update (records session changes) — neither does backlog cleanup.
Run cadence: whenever changelog-ai.md feels long, when resolved issues clutter known-issues.md, or at the start of a new project phase.