fix(tui): transcript copy polish batch (#4142 #4143 #4144 #4145)

[Hmbown]

Summary

Copy and dedup polish for the v0.8.68 "transcript copy polish" lane. Low-risk wording/dedup only — no user-facing rendering-logic changes.

Fixed

• v0.8.68 5.2: trim redundant mode picker body copy #4143 — trim redundant mode picker body copy. ModePickerPrompt ("Choose how CodeWhale should operate:") is no longer rendered by the picker view — the Mode title carries the intent — so it was dead, duplicated copy across all seven locales plus a parity test. Removed the orphaned MessageId, its ALL_MESSAGE_IDS entry, the translation-test row, and the string from every locale pack.
• v0.8.68 5.4: consolidate repeated setup hints #4144 — consolidate repeated setup hints. The six "Enter records this X snapshot" review hints now share one uniform pattern (stem first, then the step-specific action) so they read as one voice instead of six. Reworded the Hotbar, Remote, and Runtime-posture hints in en.json and updated the two detail-line tests that asserted the old order.
• v0.8.68 5.7: fix read done Searching verb mismatch #4145 — fix "read done / Searching" verb mismatch. A grep grouped under the Workspace exploration card rendered as read done · Searching for …. The card now picks the find verb when its entries are all searches, so the completed action agrees with the label (find done · Searching for …); read/list or mixed cards keep the neutral read verb. Added regression tests.

Skipped

• Refs v0.8.68 5.1: disclose context percent once #4142 — skipped. Making a normal screen show context percent "once" requires coordinated sidebar context-bar rendering logic (hide the sidebar bar when the header already discloses percent), i.e. non-trivial cross-component rendering rather than a string/dedup change. That overlaps the sidebar/transcript rendering workstream, so this issue is left untouched (label retained) to avoid rendering conflicts. The header already shows context percent by default.

Fixes #4143
Fixes #4144
Fixes #4145

Verification

• cargo fmt --all --check — clean
• cargo clippy --workspace --all-features --locked -- -D warnings -A clippy::uninlined_format_args -A clippy::too_many_arguments -A clippy::unnecessary_map_or -A clippy::collapsible_if -A clippy::assertions_on_constants — clean
• cargo test --workspace --locked — 5982 passed, 2 pre-existing environment-specific failures unrelated to this change and reproducible in isolation on an unmodified tree: git_repo_root_reports_attempted_paths_when_no_repo_found (the checkout is nested inside a real git repo, so the "no repo" assumption fails) and skill_hotbar_action_activates_skill_through_dollar_alias (depends on the machine's startup skill cache). Neither touches any file in this PR.
• cargo build --release -p codewhale-tui — clean

🤖 Generated with Claude Code

https://claude.ai/code/session_0172k8wqL9he1teotxTS5gbu

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[claude]

Claude finished @Hmbown's task in 1m 37s —— View job

---

Review complete

• Gather context (PR body, diff, changed files)
• Read full diff against origin/main
• Review correctness (history.rs verb-inference logic, tests)
• Review provider/model/route safety implications (N/A — no model/provider code touched)
• Review reuse/simplification and dead-code removal correctness (localization.rs, locale JSON)
• Review test coverage
• Post final review summary

Summary

This is a clean, low-risk copy/dedup batch as described. Verified each fix:

#4143 — ModePickerPrompt removal: Confirmed MessageId::ModePickerPrompt has zero remaining references anywhere in crates/tui/src/** after the removal, and the string is dropped consistently from all 7 locale files (en, es-419, ja, pt-BR, vi, zh-Hans, zh-Hant) plus ALL_MESSAGE_IDS and the mode_picker_strings_are_translated_in_non_english_locales parity test. No dangling references, no orphaned translations left behind.

#4144 — setup hint consolidation: The three en.json hint strings now share the "Enter records this X snapshot. Press Y to …" pattern. The corresponding setup/mod.rs unit tests (hotbar, remote) were updated in lockstep with the new wording. Good — no test/string drift.

#4145 — verb mismatch fix (crates/tui/src/tui/history.rs:838-844,1718-1733): The new exploring_card_family helper checks whether every ExploringEntry.label starts with "Searching" to pick ToolFamily::Find vs ToolFamily::Read. Note this label is the fixed English string produced by exploring_label() in tool_routing.rs (not a localized MessageId), so the string match is stable across all locales — this isn't a locale-fragility bug.

The change correctly reuses the existing render_tool_header_with_family_and_summary (already used by GenericToolCell) instead of introducing new rendering logic — good reuse, no duplication.

Minor test-coverage gap: the two new tests (exploring_card_search_reads_as_find_not_read, exploring_card_read_keeps_read_verb) cover the pure-search and pure-read cases, but not the mixed case (a search entry alongside a read/list entry in the same card) — which is the scenario the doc comment and PR description explicitly call out as staying on the read verb ("read/list or mixed cards keep the neutral read verb"). Worth a follow-up test asserting a mixed-entry ExploringCell still renders read done, since that's the one branch of exploring_card_family's all_search logic that isn't currently exercised.

No correctness bugs, no provider/model routing concerns (no such code touched), no security issues, no dead code left over from the removals.

Verification note: I was not able to run cargo test/cargo build in this environment (Bash tool execution required approval that wasn't available), so I relied on manual diff/reference-tracing review instead of re-running the test suite the PR description already reports as passing (5982 passed, 2 pre-existing unrelated failures).