UI and snapshot verification
Issue #25, verification run on 2026-10-12 against loop/issue-25 at 2393e75. Testing-plan rows: T5, T6.
TUI
uv run pytest tests/ui
text
9 snapshots passed.
27 passed in 34.01s
rc=0 secs=37These are the SVG snapshots of the new states, from the committed snapshot tests:
input-required-with-sub-state:slash-completion-and-hint:sub-task-row:sub-task-open:completed-state:
Web
cd renderers/web && bun run test:visual (Playwright, system Chromium via TINY_HARNESS_CHROMIUM)
text
12 passed (23.4s)
rc=0 secs=36controls-light:
controls-dark:
controls-phone:
listbox:
sub-task-created:
sub-task-open:
completed-light:
completed-dark:
completed-phone:
The flow goes: reply → / listbox → /sub-task → open it → Complete the sub-task → ← Parent → Complete.

The screens match the locked prototypes in docs/specs/issue-25/design/, except where design.md changed during implementation:
- Web sub-task rows show the id and the agent, not "state · goal": showing state and goal would cost an extra request per sub-task.
- In the TUI, the reply is drawn before its state line.