factory-memory-cache

Memory caching accelerates build automation by eliminating redundant file I/O overhead when agents repeatedly access the same project artifacts across sequential commands. It stores feature states and execution plans in a unified repo cache for cross-command reuse.

To cache feature states and execution plans, leverage the `/memories/repo/` system to store frequently accessed data like build commands and codebase inventories. This enables cross-command reuse so sequential operations retrieve cached data instead of re-parsing files.

Write-through caching guarantees data consistency for codebase inventories by implementing source hash validation to ensure cached data never goes stale. It also features graceful degradation, falling back to direct file reads if the cache is unavailable or corrupted.

You can use memory caching for large project build preparation in Myrmion AI Factory to cut preparation time significantly. When running multiple IMPLEMENT --build commands, the cache eliminates repeated parsing of the codebase inventory and governance snapshot.

Unified repo caching enforces a 500-line cache size limit to ensure data consistency and non-blocking command execution. When the cache fails or becomes corrupted, the system gracefully degrades by falling back to direct file reads.