Before submitting
Area
apps/server
Problem or use case
t3.json can define an action with runOnWorktreeCreate: true, but it does nothing until each user manually imports that action. I miss Codex's code-first setup flow, where repository setup travels with the code.
Proposed solution
Let the worktree setup runner use the checked-in action directly from t3.json. This could be gated by repository trust or a managed opt-in, with approval requested again if the command changes.
Why this matters
Teams could keep setup in one checked-in place instead of relying on every teammate to repeat UI configuration or discover that their worktree was not initialized.
Smallest useful scope
When there is no saved setup action, load the runOnWorktreeCreate action from t3.json after the repository has been trusted.
Alternatives considered
Manual import is easy to miss and can drift from the file. Agent instructions run too late, and Git hooks require separate setup; #551 also notes that their output is currently hidden.
Risks or tradeoffs
Running repository code is a trust boundary, so automatic setup should require explicit repository trust or an administrator-controlled policy.
Examples or references
Related: #551, which covers Git hook output rather than using t3.json directly.
Contribution
Before submitting
Area
apps/server
Problem or use case
t3.jsoncan define an action withrunOnWorktreeCreate: true, but it does nothing until each user manually imports that action. I miss Codex's code-first setup flow, where repository setup travels with the code.Proposed solution
Let the worktree setup runner use the checked-in action directly from
t3.json. This could be gated by repository trust or a managed opt-in, with approval requested again if the command changes.Why this matters
Teams could keep setup in one checked-in place instead of relying on every teammate to repeat UI configuration or discover that their worktree was not initialized.
Smallest useful scope
When there is no saved setup action, load the
runOnWorktreeCreateaction fromt3.jsonafter the repository has been trusted.Alternatives considered
Manual import is easy to miss and can drift from the file. Agent instructions run too late, and Git hooks require separate setup; #551 also notes that their output is currently hidden.
Risks or tradeoffs
Running repository code is a trust boundary, so automatic setup should require explicit repository trust or an administrator-controlled policy.
Examples or references
Related: #551, which covers Git hook output rather than using
t3.jsondirectly.Contribution