Neat. It looks a bit more like agent swarm management from the README?
gavel is designed more for sanity checking before commits, or discussing larger technical project plans where you have multiple issues raised concurrently.
I really haven't had much luck with "do a big thing, go for it" prompts. I think it's way too easy, at least for me, for human/robot disconnect to happen, where I *think* we're on the same page, and then it builds something quite wrong.
To overcome that, I've been using "technical project plans" to drive agents -- https://photostructure.com/coding/claude-code-tpp/ -- the agent produces a doc that I can read and vet and edit, and _then_ they can go cook. I use gavel to discuss the TPP before they cook, if it's tricky.
[0]: https://plannotator.ai/
gavel is designed more for sanity checking before commits, or discussing larger technical project plans where you have multiple issues raised concurrently.
I really haven't had much luck with "do a big thing, go for it" prompts. I think it's way too easy, at least for me, for human/robot disconnect to happen, where I *think* we're on the same page, and then it builds something quite wrong.
To overcome that, I've been using "technical project plans" to drive agents -- https://photostructure.com/coding/claude-code-tpp/ -- the agent produces a doc that I can read and vet and edit, and _then_ they can go cook. I use gavel to discuss the TPP before they cook, if it's tricky.