The memwal_remember MCP tool's description states "Call ONLY when the user explicitly asks to remember/save something." This directly conflicts with building any agent designed to write memory proactively based on its own judgment (e.g. on decisions, failures, or errors) rather than on explicit user request — which is a core, encouraged pattern for agent memory systems. Recommend softening this guardrail or documenting an explicit opt-in for proactive-write agents.
The memwal_remember MCP tool's description states "Call ONLY when the user explicitly asks to remember/save something." This directly conflicts with building any agent designed to write memory proactively based on its own judgment (e.g. on decisions, failures, or errors) rather than on explicit user request — which is a core, encouraged pattern for agent memory systems. Recommend softening this guardrail or documenting an explicit opt-in for proactive-write agents.