Using the custom slash command functionnality

[Gabbo7474]

It would be interesting to see if some of the tasks could be extracted as custom slash commands.
https://docs.anthropic.com/en/docs/claude-code/tutorials#create-custom-slash-commands

[cabinlab]

Ha! I logged in to start a conversation on slash commands!!!

Here's where I was going with it: I started experimenting with parallelization. This quickly lead to thinking about token use and/or rate-limiting (depending on API vs. Max subscription). The obvious next step was considering Opus vs. Sonnet modes for parallel tasks. That lead to the /model command. What I wasn't able to find was whether it's possible to add flags to that command to allow switching without requiring human intervention (the standard command pops up a selection process).

Where it got weird was discovering Claude is unable to see any references to the model switching in the CLI docs. Struck me as super odd, and seems like Anthropic is doing that intentionally.

I think it also makes sense to adapt Claude Code's native memory functions.

Your suggestion to also use custom slash commands is smart, and I'll definitely look into it.

TBH, I predict this will all significantly diverge from BMAD soon. Claude Code always tries to remove and redesign the structure. I want to be respectful of their work, so calling this BMAD as it diverges from their vision may confuse and detract - which I don't want to do.

My current thinking is to keep this fork as close to the BMAD method as possible, and build a Claude Code method from the ground up in a new repo.

Those tangents aside, did you have specifics in mind regarding custom slash commands?

[Gabbo7474]

Very nice! Do you plan on making it public when you think it's ready or will you keep it a personal project?

[cabinlab]

Definitely plan on opening it up. Felt awkward with the BMAD fork because people found it and started using it when I was still just messing around, so I felt a responsibility to not break what they were doing while trying to "finish" it.

Claude thinks the new version is ready to launch but 80% done. I think it will take one more day of productive work to make it public with caveats.

[cabinlab]

Recent Additions

• Built in dynamic thinking (hard, ultrathink, etc.) adjustment
• Model switching to reserve overhead for flexible Opus use and/or token optimization
• Continuous running background agent (Haiku) to monitor conflicts

[Gabbo7474]

How do you run haiku for continuous monitoring in the background?

[cabinlab]

Terminal parallelization (headless)