Thank you for sharing! Wow!! I built a very similar skill for my self a few months ago :) I use a different naming convention but high-level we both seem to be solving the same problem. Can't believe these things aren't built-in into the coding agents out of the box. My version: https://github.com/vmysla/agent-skill-product-traceability
Really cool to see someone else solving this — the hook-driven continuous approach vs. my manual end-of-session command feel complementary rather than competing.
Name's kinda incidental — I kept typing "document, memories, git" to my Claude session at the end of every session until it just became /dmg.