alternative: Aperant (parallel agents), Mulica-ai (clear dashboard)

• WP → Work package
• FR → functional requirement
• T00x → task
• One branch per work package
• Open Questions / concerns

  1. SpecKitty does not push the worktree to GitHub once review/implement is done.

  2. Its not /specify, /review, /implement are parameterized to a WP0x, or can be run without any params (and what happens then, are all tasks resumed?) Clarity needed.

  3. Need a general command/escape hatch called /spec-kitty.do where you can write a series of commands, like “implement WP04 and review WP05” in one go, or give complex instructions that don’t fit into the workflow. Also useful for first time users.

  4. Its not clear how context is being shared between Claude in general, and the /spec commands. Does using of /spec cause context to mix with non command talk? Anthropic Code 101 actually.

  5. After every command, does SpecKitty store the result, instead of auto-running the next command. For example, when I do /review, does it merely store the review result, and only a /implement then reads from review?

  6. Whats the code review flow - after each implement, you run /review (for spec adhere check), then check the code in vscode/github, do some more /review specific, and finally do ‘accept’ and ‘merge’

  7. See preimplementation plan/task? its not possible right now. Also a slash called “next-task-plans” could open the dashboard and page dedicated to “top” of the current kanban.

     1. Popup is available, but “cursor” should be of type “pointer”

  8. Add next action as CTA’s on each task. In the Kanban view. And for review items, show the Code changes/redirect to github compare.

  9. There’s no way to delete a wrong spec. It says:

     If you want to fully wipe the existing spec first, you can just delete the file manually or overwrite it.

  10. Weird naming for “accept” (should be validate), and “merge”, should be “consolidate”. Neither does SpecKitty nudge you to consolidate once all WPs are done. I had to manually do that. Very weird.

      1. During merge, it merged in the current branch (main), instead of creating a new branch named ‘01-booking-cancellation-penalty-engine’, which I would have given for review. This is either a bug, or SpecKitty wants me to give it branch-name that it should create. Either way, its not good UX because worktrees and consolidation of their changes should by default, happen in a branch named like the spec. Without touching main. Its too complicated right now.
      2. Also WP branches aren’t autopushed, I need to keep asking SpecKitty to do so. This is bad. WP branches should not even be a construct when working.
      3. Another major failure: post consolidation, my dashboard says progress is zero.

  11. Need a “what to do next command” in SpecKitty (if the human loses working context).

      1. Also, if specKitty suggests a next move, add a shortcut key to just say “yes” or “no” instead of having to copy and paste the next move manually

         

      Problems with SDD

  12. clear visibility of code during review (create a slash command that opens on github in /compare mode). Its hard to see what code changes were made for a worktree, this should be easily viewable. on GitHub or on vscode.

  13. auto spawning multiple agents. Aperant already does this https://github.com/AndyMik90/Aperant (YOLO mode)

  14. see the work progress as a DAG whose nodes are clickable (implement, review are buttons, with progress viewable too).

  15. After implement, and review, does spec-kitty push the WP branches? its not clear, neither is there a “push all command” (State is stuck on my machine). /spec-kitty.push could be a command

  16. How to add TDD into this, superpowers?

graph TD
  research --> specify --> clarify --> constituion --> plan --> tasks --> implement_X
  clarify --> plan
  plan --> plan
  clarify --> clarify
  implement_X --> PI --> accept --> merge
  subgraph PI[after implementation]
    checklist
    review_x
		analyze
		view(/compare)
  end
  
  status
  dashboard