## What changed
Reject `shell_command` and unified `exec_command` calls that provide a
`justification` without also specifying `sandbox_permissions`. The
model-visible error tells callers to explicitly request `require_escalated`
execution or omit the justification.
## Testing
- Add unit coverage for `shell_command` argument validation.
- Add integration coverage for both shell tool paths, including verification
that rejected unified exec calls do not begin execution.
GitOrigin-RevId: 6af39d2c889af492440615eae12085b99a8b5169
Fixes#27210.
## Why
When the app server reports a visible `HookStarted` event for a
`PostToolUse` hook but the turn reaches `TurnCompleted` before a
matching hook completion event arrives, the TUI can leave the transient
`Running PostToolUse hook` row visible after the agent is done.
Interrupted and failed turn cleanup already drops transient live hook
rows; the normal completion path did not.
## What Changed
- Added `ChatWidget::clear_active_hook_cell()` for dropping transient
live hook status without writing it to history.
- Call that cleanup from normal task completion, while reusing it for
the existing start/finalize cleanup paths.
- Added `completed_turn_clears_visible_running_hook` snapshot coverage
for the reported `PostToolUse` case.
## Tests
- `just test -p codex-tui completed_turn_clears_visible_running_hook`
- `just test -p codex-tui` (fails on current `main` in unrelated
guardian tests:
`update_feature_flags_disabling_guardian_clears_review_policy_and_restores_default`
and
`update_feature_flags_disabling_guardian_clears_manual_review_policy_without_history`)