clockworksoul/enso ? reverse-engineered prompt

Build me the first working version of Ensō, a portable memory system for AI agents where Markdown is the source of truth and everything else is just a rebuildable index.

I want the core to let me append memories, supersede old ones without deleting anything, and rank what should be recalled later. It should work even if no framework is present, so a fresh clone of the repo and the Markdown files alone should be enough to recover state. Please keep the design fail safe, so if something breaks it falls back to the current simple behavior instead of blocking use.

Start with the internal core and the Markdown store, then make sure the command line tools can add and validate entries cleanly. If the graph backed store is already in the repo, keep it compatible with the same store interface, but do not overbuild anything speculative. Make the round trip from Markdown to memory and back lossless, and make the tests prove that the append only and rebuildable behavior works. If you need to check current docs or library details online, go ahead and do that.