Feature request: /memories command to show loaded memory files

Problem

There is no way for users to verify which memory files have been loaded into the current conversation context. The memory system relies on Claude self-reporting whether it read memories, which is unverifiable and unreliable due to non-deterministic behavior.

MEMORY.md is injected into the system context, but whether individual memory files referenced in the index were actually read is opaque to the user.

Proposed Solution

Add a /memories command (or similar) that lists:

• Which memory files exist in the project memory directory
• Which ones have been loaded into the current conversation context
• Optionally, a brief summary of each (the description field from frontmatter)

This would give users a concrete way to verify memory state rather than relying on the model's self-report.

Why It Matters

Users who invest time writing memories to guide Claude's behavior have no feedback loop to confirm those memories are actually being used. This erodes trust in the memory system and makes debugging missed instructions difficult.

A simple introspection command would close that gap.