akitaonrails/ai-memory ? reverse-engineered prompt

Build me a Rust tool that gives AI coding agents long term memory across sessions and even across different agent apps.

I want it to watch the agent lifecycle, capture the important parts of a task, and turn that into a clean markdown wiki that lives in a git repo so it is easy to search, open in Obsidian, back up, and share. When one session ends, it should create a short handoff that the next agent can pick up without me having to explain the same architecture, dead ends, or open questions again.

It should work with common coding agents through MCP and hooks, and support a managed run mode that can resume a workstream in the same project. Keep the captured context sanitized and bounded, and make sure it ignores sensitive or noisy stuff.

I mainly want something that works well on Linux, macOS, and in WSL, with clear install and setup commands, and if you need current docs for any agent integration, look them up online.