feat(i18n): extract core_commands.py user-facing strings

[thomwebb]

What

Migrates 41 user-facing emit_* strings across 7 slash-command handlers in command_line/core_commands.py to the t() catalog — the #2 extraction target at 47 raw audit sites (41 genuinely translatable; 6 false positives).

Result

47 raw → 6 — all remaining hits are false positives:

• Rich-markup f-strings in picker-failure fallback paths (Text.from_markup(f"[dim]...{var}...[/dim]"))
• Traceback diagnostics (f"Traceback: {traceback.format_exc()}")
• Pure-variable agent description strings (f"{new_agent.description}")

31 new cmd.* keys added to en-US.json; 5 reuse existing keys:

Existing key reused	From
cli.goodbye	/exit goodbye message
cfg.agent.not_found	agent not-found error
cli.agent.available	agent listing
cmd.agent.picker_failed	model picker failure (same message)

Key groupings

Namespace	Commands covered
cmd.cd.*	/cd: success, list-error, agent-updated, reload-error, not-a-dir
cmd.paste.*	/paste: no-image, hint, count, send-hint, failed
cmd.tutorial.*	/tutorial: chatgpt-oauth, claude-oauth, complete, skipped
cmd.agent.*	/agent: already-using, switch-failed, switched, cancelled, picker-failed, usage
cmd.model.*	/model: success, cancelled, usage, available
cmd.add_model.*	/add_model: success, failed
cmd.model_settings.*	/model_settings: success, agent-reloaded, reload-failed, failed
cmd.plan.usage	/plan: usage error

Bonus fix

Renamed the loop variable t → token in handle_generate_pr_description_command — it was shadowing the module-level t() import (caught by ruff F402).

Tests

tests/i18n/test_core_commands_i18n.py (10 tests): namespace populated, all keys resolve + pseudolocalize, interpolation for each command group, no leftover placeholders, module imports cleanly. All 128 existing i18n tests still pass. ruff clean.

Independence

Touches only core_commands.py, en-US.json, and the new test — no overlap with any open PR.

[thomwebb]

Adversarial Review — Automated (code-puppy-4aeaa3)

Verdict: CONDITIONAL PASS — mechanical extraction is correct, no crash vectors. Three issues before merge.

---

Must Fix

1. Key reuse: cmd.agent.picker_failed used inside handle_model_command

core_commands.py:488 (approx) — the model command's picker-failure path reuses cmd.agent.picker_failed. If a translator ever needs to word these differently (agent picker vs model picker), there's no seam. Create cmd.model.picker_failed in en-US.json and use it in handle_model_command.

2. Test namespace floor too loose

tests/i18n/test_core_commands_i18n.py — assert len(_cmd_keys()) >= 25 allows 6 keys to silently vanish. The PR adds 31 keys; floor should be >= 31.

3. No targeted interpolation test for cmd.add_model.failed

The key has an {error} placeholder. test_no_leftover_placeholder_for_supplied_params catches this generically with "X", but a param-name typo in the catalog would still pass. Add one explicit assert.

---

Track / Follow-up (not merge blockers)

• Text.from_markup("[yellow]Usage:[/yellow] /agent <agent-name>") near line 384 is content-identical to cmd.agent.usage but bypasses t() — will drift if a translator updates the catalog key
• "No description" fallback (line 371) is user-visible, still raw
• "Failed" in result dispatch in the undo handler will silently break if undo_manager.py ever gets localized — track in a ticket
• Add # i18n: skip - developer output comments on the two Traceback: f-string lines so future auditors don't re-flag them

---

Everything else checks out: import is at module level, all 31 keys exist in en-US.json, all param names match, t() is crash-safe (missing key returns key, missing param leaves placeholder intact), no Rich markup in catalog values.

[thomwebb]

Review feedback addressed — commit 3b67f7ed

Automated fix pass (code-puppy-4aeaa3). Applied the three "must fix" items:

Landed:

• Added new key cmd.model.picker_failed ("Model picker failed to open: {error}"); handle_model_command no longer reuses cmd.agent.picker_failed
• Test floor tightened: >= 25 → >= 32 (matches actual current key count)
• Added test_add_model_failed_interpolates_error_param — explicit interpolation check that catches param-name typos independently of the generic sweep test
• Also added the new cmd.model.picker_failed key to test_model_keys_interpolate

Verification:

• pytest tests/i18n/test_core_commands_i18n.py -q → 11 passed
• ruff check --fix / ruff format → clean
• Catalog JSON parses; 32 cmd.* keys

Deferred (track as follow-up, not blocking):

• Text.from_markup("[yellow]Usage:[/yellow] /agent <agent-name>") at ~line 384 duplicates cmd.agent.usage content but bypasses t()
• "No description" fallback at line 371 still raw
• "Failed" in result dispatch in undo handler will break if undo_manager.py gets localized
• Traceback: f-string lines still need # i18n: skip - developer output comments

CI should re-run on the new commit.

[thomwebb]

Rebased + piggyback test fix

Rebased onto latest main (now at f0dab866) and added a small piggyback commit fixing an unrelated broken test that CI was tripping over.

Context

Upstream commit f0dab866 (feat: add GPT-5.6 reasoning context and mode) added reasoning_context and reasoning_mode to supported_settings for gpt-5.6* models in chatgpt_oauth/utils.py but did not update tests/plugins/test_chatgpt_oauth_utils.py::test_add_models_to_extra_config_gpt54_and_newer_support_xhigh, which asserts a hardcoded 3-item list. Result: main's own Build and Publish to PyPI run is failing for the same reason.

What the piggyback commit does

Splits the assertion so codex-gpt-5.6-* variants expect the extra fields:

expected_settings = ["reasoning_effort", "summary", "verbosity"]
if model_name.startswith("codex-gpt-5.6"):
    expected_settings.extend(["reasoning_context", "reasoning_mode"])
assert model_config["supported_settings"] == expected_settings

Verified locally: test passes on this branch.

Cleanup

This commit is a piggyback and will drop from the branch automatically once the same fix lands on upstream main via a rebase.