fix(tests): fix BDD step setup for issue #4741 @token extraction tests

- Replace __import__() with importlib.import_module() so step code
  correctly accesses cleveragents.tui.app (not the top-level package)
- Fix context.add_cleanup(_restore_modules) → lambda so context arg
  is passed when Behave invokes the cleanup
- Remove redundant double-quotes from Examples table values; the step
  definition's own quotes already delimit the parameter, causing the
  regex to capture a trailing quote and the assertion to fail
- Apply ruff format to bring file in line with project style

ISSUES CLOSED: #4741

CHANGELOG.md

@@ -7,6 +7,13 @@ Changed `wf10_batch.robot` to be less likely to create files, and
 
 ## [Unreleased]
 
+- **fix(tui): extract @token text correctly in on_input_submitted
+  suggestions query (PR #11004 / issue #4741)** — The TUI handler passed
+  ``text.replace("@", "").strip()`` to ``suggestions()``, producing garbage
+  fuzzy matches (e.g. "analyse \@proj" → "analyse proj"). Replaced with
+  ``re.findall(r"@(\S+)", text)`` that extracts only the last ``@token`` as
+  the query, returning just the token value without surrounding words.
+
 - **SubplanExecutionService lazy wiring in CLI** (#10268): Wired `subplan_service`
   from the DI container into `_get_plan_executor()` so `PlanExecutor` can spawn child
   plans during the Execute phase. When `SubplanService` is available but

CONTRIBUTORS.md

@@ -48,3 +48,4 @@ Below are some of the specific details of various contributions.
 * HAL 9000 has contributed the DecisionService wiring for PlanExecutor strategize persistence fix (#10813): added decision_service to the PlanExecutor constructor and wired it from the CLI dependency-injection container in `_get_plan_executor()`, plus implemented `_persist_strategy_decisions()` to persist strategy decisions as domain `Decision` objects.
 * HAL 9000 has contributed the A2A module rename standardization BDD tests (PR #10583 / issue #8615): comprehensive Behave test suite validating that all 22 A2A symbols are properly exported from `cleveragents.a2a`, no legacy ACP references remain in the module source, and documentation uses correct A2A naming conventions — fixing inline imports, unused behave symbols, cross-scenario context dependencies, and missing type annotations.
 * HAL 9000 has contributed the `ActorSelectionOverlay._render` → `_refresh_display` rename fix (PR #11176 / issue #11039, Epic #8174): renamed `_render()` method to `_refresh_display()` to avoid shadowing Textual's `Widget._render()`, fixing a crash in textual >=1.0 where `get_content_height()` would receive `None` and raise `AttributeError: 'NoneType' object has no attribute 'get_height'`.
+* HAL 9000 has contributed the TUI @token text extraction fix for suggestions query (PR #11004 / issue #4741): replaced ``text.replace("@", "").strip()`` with ``re.findall(r"@(\S+)", text)[-1]`` in ``on_input_submitted`` to correctly extract only the last @token as the query, preventing garbage fuzzy matches. Includes TDD BDD regression test suite covering single-token, category-prefixed, multi-token, and standalone-token scenarios (PR #11004).

features/steps/tdd_tui_suggestions_query_extraction_4741_steps.py

@@ -0,0 +1,252 @@
+"""Step definitions for TDD issue #4741: @token text extraction for suggestions().
+
+Verifies that ``on_input_submitted`` in ``src/cleveragents/tui/app.py`` extracts
+only the last ``@token`` from input using regex (``re.findall(r"@(\\S+)", text)``),
+instead of the buggy ``text.replace("@", "").strip()``.
+
+Uses a Scenario Outline so each <input> / <expected_query> pair is tested identically.
+"""
+
+from __future__ import annotations
+
+import contextlib
+import importlib
+import shutil
+import sys
+import tempfile
+from pathlib import Path
+from types import ModuleType, SimpleNamespace
+from typing import Any
+from unittest.mock import MagicMock
+
+from behave import given, then
+
+
+# ---------------------------------------------------------------------------
+# Mock Textual
+# ---------------------------------------------------------------------------
+
+_MOCK_TEXTUAL_KEYS = [
+    "textual",
+    "textual.app",
+    "textual.containers",
+    "textual.widgets",
+]
+
+
+def _build_mock_textual() -> dict[str, ModuleType]:
+    mock_textual = ModuleType("textual")
+    mock_app_mod = ModuleType("textual.app")
+    mock_containers = ModuleType("textual.containers")
+    mock_widgets = ModuleType("textual.widgets")
+
+    class MockApp:
+        def __init__(self, *a, **kw):
+            pass
+
+    class MockVertical:
+        def __enter__(self):
+            return self
+
+        def __exit__(self, *a, **kw):
+            pass
+
+    class MockHeader:
+        def __init__(self, *args, **kwargs):
+            pass
+
+    class MockFooter:
+        def __init__(self, *args, **kwargs):
+            pass
+
+    class MockStatic:
+        def __init__(self, *a, **kw):
+            self._text = ""
+
+        def update(self, text):
+            self._text = str(text)
+
+    class MockInput:
+        value: str = ""
+
+        def __init__(self, *a, **kw):
+            pass
+
+        def consume_text(self):
+            return SimpleNamespace(text=self.value)
+
+    mock_app_mod.App = MockApp
+    mock_containers.Vertical = MockVertical
+    mock_widgets.Header = MockHeader
+    mock_widgets.Footer = MockFooter
+    mock_widgets.Static = MockStatic
+    mock_widgets.Input = MockInput
+
+    return {
+        "textual": mock_textual,
+        "textual.app": mock_app_mod,
+        "textual.containers": mock_containers,
+        "textual.widgets": mock_widgets,
+    }
+
+
+def _restore_modules(context):
+    for key, val in getattr(context, "_tui_saved_modules", {}).items():
+        if val is None:
+            sys.modules.pop(key, None)
+        else:
+            sys.modules[key] = val
+
+    importlib.reload(
+        importlib.import_module("cleveragents.tui.widgets.help_panel_overlay")
+    )
+    importlib.reload(importlib.import_module("cleveragents.tui.widgets.persona_bar"))
+    importlib.reload(importlib.import_module("cleveragents.tui.widgets.prompt"))
+    importlib.reload(
+        importlib.import_module("cleveragents.tui.widgets.reference_picker")
+    )
+    importlib.reload(
+        importlib.import_module("cleveragents.tui.widgets.slash_command_overlay")
+    )
+    importlib.reload(importlib.import_module("cleveragents.tui.app"))
+
+
+# ---------------------------------------------------------------------------
+# Given: set up mock + run on_input_submitted with the captured input text
+# ---------------------------------------------------------------------------
+
+
+@given('the TUI on_input_submitted handler processes "{input_text}"')
+def step_run_on_input_submitted(context, input_text):
+    """Install mock Textual, create app, simulate submission, capture query.
+
+    We patch both ``ReferencePickerOverlay.set_suggestions`` and the module-level
+    ``suggestions()`` so we can inspect the exact argument passed to it.
+    """
+    # Remove saved modules from any previous scenario (Scenario Outline)
+    _restore_modules(context) if getattr(context, "_tui_saved_modules", None) else None
+
+    mocks = _build_mock_textual()
+
+    context._tui_saved_modules = {}
+    for key in _MOCK_TEXTUAL_KEYS:
+        context._tui_saved_modules[key] = sys.modules.pop(key, None)
+    for key, mod in mocks.items():
+        sys.modules[key] = mod
+
+    # Reload widget modules with mocked base classes
+    from cleveragents.tui.widgets import (
+        help_panel_overlay as hp_mod,
+        persona_bar as pb_mod,
+        prompt as prompt_mod,
+        reference_picker as rp_mod,
+        slash_command_overlay as sco_mod,
+    )
+
+    importlib.reload(hp_mod)
+    importlib.reload(pb_mod)
+    importlib.reload(prompt_mod)
+    importlib.reload(rp_mod)
+    importlib.reload(sco_mod)
+
+    # Install mock persona state and command router before app module reloads
+    tmp_dir = tempfile.mkdtemp()
+    context._tui_tmpdir = Path(tmp_dir)
+
+    from cleveragents.tui.persona.registry import PersonaRegistry
+    from cleveragents.tui.persona.state import PersonaState
+
+    registry = PersonaRegistry(config_dir=Path(tmp_dir))
+    registry.ensure_default()
+    persona_state = PersonaState(registry=registry)
+
+    class MockCommandRouter:
+        def handle(self, raw, *, session_id):
+            return f"handled:{raw}"
+
+    mock_router = MockCommandRouter()
+
+    context._tui_persona_state = persona_state
+    context._tui_cmd_router = mock_router
+
+    app_mod = importlib.import_module("cleveragents.tui.app")
+    importlib.reload(app_mod)
+
+    # Patch the suggestions function BEFORE instantiating the app, so that when
+    # on_input_submitted calls it internally, our spy intercepts it.
+    captured_queries: list[str] = []
+
+    def _spy_suggestions(query, *, category=None, limit=8):
+        captured_queries.append(query)
+        return list(
+            app_mod.suggestions.__wrapped__.__code__.co_varnames[:limit]
+        )  # dummy
+
+    # The suggestions function is imported at module level in app.py. We need to
+    # patch it there, not on reference_parser itself.
+    def spy(query, **kw):
+        captured_queries.append(query)
+        return []
+
+    app_mod.suggestions = spy
+
+    cls = app_mod._ResolvedTuiApp
+
+    # Build mock widgets dict so query_one works
+    app_inst = cls(command_router=mock_router, persona_state=persona_state)
+
+    # Manually populate _widgets from the MockApp instance if it exists
+    if hasattr(app_inst, "_widgets"):
+        wdict: dict[str, Any] = app_inst._widgets
+    else:
+        wdict = {}
+
+    # Create mock widgets that behave correctly
+    mock_prompt = MagicMock()
+    mock_prompt.consume_text.return_value = SimpleNamespace(text=input_text)
+
+    mock_picker = MagicMock()
+    mock_picker.set_suggestions = lambda q, s: None  # dummy, we spy suggestions instead
+    wdict["#reference-picker"] = mock_picker
+
+    mock_prompt_widget = MagicMock()
+    mock_prompt_widget.consume_text.return_value = SimpleNamespace(text=input_text)
+    wdict["#prompt"] = mock_prompt_widget
+
+    wdict["#conversation"] = MagicMock()
+    wdict["#help-panel"] = MagicMock()
+    wdict["#persona-bar"] = MagicMock()
+
+    # Patch query_one to use our widget dict
+    app_inst.query_one = lambda sel, *args, **kw: wdict.get(sel, MagicMock())
+
+    with contextlib.suppress(Exception):
+        app_inst.on_input_submitted(event=MagicMock())
+
+    context._tui_captured_query = captured_queries[0] if captured_queries else None
+
+    def _cleanup():
+        if hasattr(app_inst, "_widgets"):
+            app_inst._widgets.clear()
+
+    def _cleanup_tmpdir(d=context._tui_tmpdir):
+        if d is not None and d.exists():
+            shutil.rmtree(str(d), ignore_errors=True)
+
+    context.add_cleanup(_cleanup)
+    context.add_cleanup(lambda: _restore_modules(context))
+    context.add_cleanup(_cleanup_tmpdir)
+
+
+# ---------------------------------------------------------------------------
+# Then: assertion on the captured query
+# ---------------------------------------------------------------------------
+
+
+@then('the suggestion query should be "{expected}"')
+def step_suggestion_query_is_correct(context, expected):
+    assert context._tui_captured_query == expected, (
+        f"Bug #4741: extracted suggestion query was "
+        f"{context._tui_captured_query!r}, expected {expected!r}. "
+        f"The old buggy .replace(@).strip() approach produces garbage matches."
+    )

features/tdd_tui_suggestions_query_extraction_4741.feature

@@ -0,0 +1,23 @@
+@tdd_issue @tdd_issue_4741
+Feature: TDD Issue #4741 — Extract only the @token text as suggestions() query
+
+  Bug #4741: ``on_input_submitted`` passed ``text.replace("@", "").strip()`` to
+  ``suggestions()``, e.g. "analyse \@proj" → "analyse proj", producing garbage
+  fuzzy matches because surrounding words polluted the query.
+
+  The fix replaces it with ``re.findall(r"@(\S+)", text)`` taking the last
+  match as the query string, so "analyse \@proj" → "proj".
+
+Scenario Outline: @token extraction produces correct queries for input text
+  The suggestion query must be the last extracted @token from input text.
+  With the old buggy code these assertions fail; after the fix they pass.
+
+  Given the TUI on_input_submitted handler processes "<input>"
+  Then the suggestion query should be "<expected_query>"
+
+  Examples:
+    | input                       | expected_query      |
+    | analyse @proj               | proj                |
+    | read @project:main-doc      | project:main-doc    |
+    | look-up @actor:x @tool:y   | tool:y              |
+    | @skill:doc-generator        | skill:doc-generator |

src/cleveragents/tui/app.py

@@ -4,6 +4,7 @@ from __future__ import annotations
 
 import importlib
 import os
+import re
 from dataclasses import dataclass, field
 from typing import TYPE_CHECKING, Any, ClassVar, Protocol
 
@@ -396,10 +397,13 @@ if _TEXTUAL_AVAILABLE:
 
             expanded = result.expanded_text
             if "@" in text:
-                ref_picker = self.query_one("#reference-picker", ReferencePickerOverlay)
-                ref_picker.set_suggestions(
-                    text, suggestions(text.replace("@", "").strip())
-                )
+                matches = re.findall(r"@(\S+)", text)
+                if matches:
+                    query_text = matches[-1]
+                    ref_picker = self.query_one(
+                        "#reference-picker", ReferencePickerOverlay
+                    )
+                    ref_picker.set_suggestions(text, suggestions(query_text))
 
             if self._facade is None:
                 # Facade not wired yet — preview only (graceful degradation)