Merge PR #293: fix: eliminate shell noise from terminal output and fix test failures

Authored by 0xbyt4. Wraps commands with unique fence markers to isolate real output
from shell init/exit noise (oh-my-zsh, macOS session restore, etc.). Falls back to
expanded pattern-based cleaning. Also fixes BSD find fallback and test module shadowing.

tests/tools/test_file_tools_live.py

@@ -17,7 +17,13 @@ import pytest
 
 sys.path.insert(0, str(Path(__file__).resolve().parents[2]))
 
-from tools.environments.local import LocalEnvironment, _clean_shell_noise, _SHELL_NOISE_SUBSTRINGS
+from tools.environments.local import (
+    LocalEnvironment,
+    _clean_shell_noise,
+    _extract_fenced_output,
+    _OUTPUT_FENCE,
+    _SHELL_NOISE_SUBSTRINGS,
+)
 from tools.file_operations import ShellFileOperations
 
 
@@ -28,6 +34,7 @@ from tools.file_operations import ShellFileOperations
 _ALL_NOISE_PATTERNS = list(_SHELL_NOISE_SUBSTRINGS) + [
     "bash: ",
     "Inappropriate ioctl",
+    "Auto-suggestions:",
 ]
 
 
@@ -125,11 +132,84 @@ class TestCleanShellNoise:
         _assert_clean(result)
 
     def test_noise_in_middle_not_stripped(self):
-        """Only LEADING noise is stripped -- noise in the middle is real output."""
+        """Noise in the middle is real output and should be preserved."""
         output = "real\nbash: no job control in this shell\nmore real\n"
         result = _clean_shell_noise(output)
         assert result == output
 
+    def test_zsh_restored_session(self):
+        output = "Restored session: Mon Mar  2 22:16:54 +03 2026\nhello\n"
+        result = _clean_shell_noise(output)
+        assert result == "hello\n"
+
+    def test_zsh_saving_session_trailing(self):
+        output = "hello\nSaving session...completed.\n"
+        result = _clean_shell_noise(output)
+        assert result == "hello\n"
+
+    def test_zsh_oh_my_zsh_banner(self):
+        output = "Oh My Zsh on! | Auto-suggestions: press right\nhello\n"
+        result = _clean_shell_noise(output)
+        assert result == "hello\n"
+
+    def test_zsh_full_noise_sandwich(self):
+        """Both leading and trailing zsh noise stripped."""
+        output = (
+            "Restored session: Mon Mar  2\n"
+            "command not found: docker\n"
+            "Oh My Zsh on!\n"
+            "actual output\n"
+            "Saving session...completed.\n"
+        )
+        result = _clean_shell_noise(output)
+        assert result == "actual output\n"
+
+    def test_last_login_stripped(self):
+        output = "Last login: Mon Mar 2 22:00:00 on ttys001\nhello\n"
+        result = _clean_shell_noise(output)
+        assert result == "hello\n"
+
+
+# ── _extract_fenced_output unit tests ────────────────────────────────────
+
+class TestExtractFencedOutput:
+    def test_normal_fenced_output(self):
+        raw = f"noise\n{_OUTPUT_FENCE}hello world\n{_OUTPUT_FENCE}more noise\n"
+        assert _extract_fenced_output(raw) == "hello world\n"
+
+    def test_no_trailing_newline(self):
+        """printf output with no trailing newline is preserved."""
+        raw = f"noise{_OUTPUT_FENCE}exact{_OUTPUT_FENCE}noise"
+        assert _extract_fenced_output(raw) == "exact"
+
+    def test_no_fences_falls_back(self):
+        """Without fences, falls back to pattern-based cleaning."""
+        raw = "bash: no job control in this shell\nhello\n"
+        result = _extract_fenced_output(raw)
+        assert result == "hello\n"
+
+    def test_only_start_fence(self):
+        """Only start fence (e.g. user command called exit)."""
+        raw = f"noise{_OUTPUT_FENCE}hello\nSaving session...\n"
+        result = _extract_fenced_output(raw)
+        assert result == "hello\n"
+
+    def test_user_outputs_fence_string(self):
+        """If user command outputs the fence marker, it is preserved."""
+        raw = f"noise{_OUTPUT_FENCE}{_OUTPUT_FENCE}real\n{_OUTPUT_FENCE}noise"
+        result = _extract_fenced_output(raw)
+        # first fence -> last fence captures the middle including user's fence
+        assert _OUTPUT_FENCE in result
+        assert "real\n" in result
+
+    def test_empty_command_output(self):
+        raw = f"noise{_OUTPUT_FENCE}{_OUTPUT_FENCE}noise"
+        assert _extract_fenced_output(raw) == ""
+
+    def test_multiline_output(self):
+        raw = f"noise\n{_OUTPUT_FENCE}line1\nline2\nline3\n{_OUTPUT_FENCE}noise\n"
+        assert _extract_fenced_output(raw) == "line1\nline2\nline3\n"
+
 
 # ── LocalEnvironment.execute() ───────────────────────────────────────────
 

tests/tools/test_terminal_disk_usage.py

@@ -5,6 +5,12 @@ from unittest.mock import patch, MagicMock
 
 import pytest
 
+# tools/__init__.py re-exports a *function* called ``terminal_tool`` which
+# shadows the module of the same name.  Use sys.modules to get the real module
+# so patch.object works correctly.
+import sys
+import tools.terminal_tool  # noqa: F401 -- ensure module is loaded
+_tt_mod = sys.modules["tools.terminal_tool"]
 from tools.terminal_tool import get_active_environments_info
 
 # 1 MiB of data so the rounded MB value is clearly distinguishable
@@ -34,10 +40,8 @@ class TestDiskUsageGlob:
             "aaaaaaaa-1111-2222-3333-444444444444": MagicMock(),
         }
 
-        with (
-            patch("tools.terminal_tool._active_environments", fake_envs),
-            patch("tools.terminal_tool._get_scratch_dir", return_value=fake_scratch),
-        ):
+        with patch.object(_tt_mod, "_active_environments", fake_envs), \
+             patch.object(_tt_mod, "_get_scratch_dir", return_value=fake_scratch):
             info = get_active_environments_info()
 
         # Task A only: ~1.0 MB. With the bug (hardcoded hermes-*),
@@ -51,10 +55,8 @@ class TestDiskUsageGlob:
             "bbbbbbbb-5555-6666-7777-888888888888": MagicMock(),
         }
 
-        with (
-            patch("tools.terminal_tool._active_environments", fake_envs),
-            patch("tools.terminal_tool._get_scratch_dir", return_value=fake_scratch),
-        ):
+        with patch.object(_tt_mod, "_active_environments", fake_envs), \
+             patch.object(_tt_mod, "_get_scratch_dir", return_value=fake_scratch):
             info = get_active_environments_info()
 
         # Should be ~2.0 MB total (1 MB per task).