prod-memory-management

To resolve workplace acronyms and project codenames for agent collaboration, use a tiered lookup approach. Check the hot working cache in CLAUDE.md first, then query deep memory stores like memory/glossary.md and project directories to decode shorthand into accurate context.

A two-tier memory system decodes internal workplace shorthand by checking a hot working cache in CLAUDE.md for immediate context, then expanding into deep memory stores under memory/ directories for richer profiles, glossary terms, and project details to ensure accurate entity resolution.

The best way to maintain a knowledge base for context decoding across changing teams is using structured directories for people, projects, and context. Apply index and log conventions to keep the hot cache and deep memory synchronized through ingest, query, and lint operations.

Yes, you can use CLAUDE.md to manage workflow instructions and partial internal language. It serves as a hot working cache that accurately resolves ambiguous directives like nicknames and acronyms into the correct people and terms before executing daily collaboration tasks.

Your agent misunderstands internal workplace shorthand and nicknames because it lacks a structured knowledge base for context decoding. Without a tiered lookup flow mapping partial internal language to correct people and projects, reliable task execution fails on ambiguous directives.