brd init suggestion

I just installed and initialized brd in one of my repos and noticed the instructions it injected into AGENTS.md was kind of long (~ 100 lines) and maybe this context doesn't need to be in every context window I open.

Might I suggest what I've noticed other agentish tools doing that seems to really be working well. I think this will obviously depend on the agent in use but taking claude as an example, does .claude/ exists and if it does mkdir .claude/skills/using-brd and write everything to a local skill @ .claude/skills/using-brd/SKILL.md. That's what I've done with the instructions locally anyway and so far so good. Maybe create a dedicated skill or plugin too that can be installed globally. if you autodetect that then nothing to be done :)

Great work, btw. I found this project when following a link you shared on an issue in the beads repo. I had a similar idea too, storing things in markdown, but you've actually gone and made the thing, which is fantastic imo. Great tui even.

Start by tracing the brd init command and how it writes AGENTS.md. Review the proposed .claude/skills/using-brd/SKILL.md location and decide how agent-specific installation should be handled. Done means initialization avoids unnecessary context in AGENTS.md while preserving the brd instructions through the agreed skill or plugin path.