context-engineering-advisor

Diagnose context stuffing by assessing symptoms like escalating token costs, vague hedged responses, and multi-step agent chains that pass everything along, then evaluate decision relevance and context boundary ownership to determine if information is shaped for reliable attention.

Multi-step agent chains produce bloated and inconsistent outputs when they suffer from context stuffing, passing all information along instead of separating what to persist vs retrieve, leading to context rot and degraded attention across steps.

Fix RAG retrieval strategy and memory architecture by defining bounded contexts, implementing a two-layer memory system, and applying a Research→Plan→Reset→Implement cycle to separate persistent data from retrieved decision-relevant information.

Reduce token efficiency issues and RAG brittleness by defining decision-relevant context boundaries, deciding what to persist vs retrieve, and fixing underlying information architecture issues to prevent context rot and improve output consistency.

You need bounded context boundaries and a Research→Plan→Reset→Implement cycle when product management workflows suffer from escalating token costs, hedging, and inconsistent outputs across multi-step agent chains and RAG memory setups.