Pause interactive launches after the role intro

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

coilyco-flight-deck/agent-compose#137

coilyco-ops commented

2026-07-30 09:40:08 +00:00

Member

Outcome

Make the assigned-role intro a deliberate, readable threshold before an interactive harness starts. Print everything at normal speed, place the role overview and ordered personality meld last, then wait for Enter before handing the terminal to the harness.

Follow-up to #136.

Current experience

#136 added the TTY-aware role transcript, but the transcript completes almost instantly and the harness starts immediately afterward. Kai manually scrolls back to read it on every launch. Claude's transition into its TUI may replace the visible terminal buffer entirely.

The current information hierarchy also puts the most interesting content, the role overview and personality meld, first. Routine composition output then pushes it away from the launch boundary.

This is a repeated launch ritual, not first-run education. Remembered “seen” state and artificial slow printing would solve the wrong problem.

Interaction

For an interactive TTY launch:

Run composition and print routine status or diagnostic output at normal speed.
Render the selected role overview and ordered personality meld as the final substantive block.
Print Press Enter to continue.
Enter starts the harness. Ctrl-C cancels without starting it.

For a non-interactive, piped, or headless launch, preserve the current behavior and never wait for input.

Acceptance criteria

Interactive launches print routine composition output before the role overview and personality meld.
The role overview and ordered meld are the final substantive content before Press Enter to continue.
The prompt remains visible until Enter is pressed, and the harness does not start earlier.
Ctrl-C at the prompt exits without launching the harness.
The gate is active only when the launch owns interactive terminal input and output.
Redirected, piped, and headless launches never block or read from stdin.
Output remains immediate. No typewriter effect, countdown, sleep, or remembered “seen” state is added.
Explicit harness arguments and the Codex introduction behavior from #136 remain unchanged.
Focused tests cover the interactive Enter path and the non-interactive no-pause path.
Native launch documentation describes the acknowledgement step and the inverted output hierarchy.

Design rationale

The extra Enter replaces an existing manual scroll-back action and guarantees the identity text survives until Kai has read it. If repeated use later shows that the acknowledgement itself is the problem, that evidence can motivate an opt-out. No preference surface is needed for this change.

## Outcome Make the assigned-role intro a deliberate, readable threshold before an interactive harness starts. Print everything at normal speed, place the role overview and ordered personality meld last, then wait for Enter before handing the terminal to the harness. Follow-up to #136. ## Current experience #136 added the TTY-aware role transcript, but the transcript completes almost instantly and the harness starts immediately afterward. Kai manually scrolls back to read it on every launch. Claude's transition into its TUI may replace the visible terminal buffer entirely. The current information hierarchy also puts the most interesting content, the role overview and personality meld, first. Routine composition output then pushes it away from the launch boundary. This is a repeated launch ritual, not first-run education. Remembered “seen” state and artificial slow printing would solve the wrong problem. ## Interaction For an interactive TTY launch: 1. Run composition and print routine status or diagnostic output at normal speed. 2. Render the selected role overview and ordered personality meld as the final substantive block. 3. Print `Press Enter to continue`. 4. Enter starts the harness. Ctrl-C cancels without starting it. For a non-interactive, piped, or headless launch, preserve the current behavior and never wait for input. ## Acceptance criteria * [ ] Interactive launches print routine composition output before the role overview and personality meld. * [ ] The role overview and ordered meld are the final substantive content before `Press Enter to continue`. * [ ] The prompt remains visible until Enter is pressed, and the harness does not start earlier. * [ ] Ctrl-C at the prompt exits without launching the harness. * [ ] The gate is active only when the launch owns interactive terminal input and output. * [ ] Redirected, piped, and headless launches never block or read from stdin. * [ ] Output remains immediate. No typewriter effect, countdown, sleep, or remembered “seen” state is added. * [ ] Explicit harness arguments and the Codex introduction behavior from #136 remain unchanged. * [ ] Focused tests cover the interactive Enter path and the non-interactive no-pause path. * [ ] Native launch documentation describes the acknowledgement step and the inverted output hierarchy. ## Design rationale The extra Enter replaces an existing manual scroll-back action and guarantees the identity text survives until Kai has read it. If repeated use later shows that the acknowledgement itself is the problem, that evidence can motivate an opt-out. No preference surface is needed for this change.

coilyco-ops referenced this issue

2026-07-30 09:43:35 +00:00

Let Designer implement visual-only web UI changes #138

coilyco-ops referenced this issue from a commit

2026-07-30 09:53:09 +00:00

feat(launch): pause for assigned identity, closes #137

coilyco-ops closed this issue

2026-07-30 09:53:10 +00:00

2026-07-30 09:56:46 +00:00

Run contract: complete the Agent Compose v1.x issue suite #115