A file called scratch.py is not the scratch directory

Running the checks against the baseline task, which has no scratch directory at all,
should report zero touches. It reported two. Both agents had made throwaway files named
scratch.py and scratch2.py beside their work, and the mention pattern's (?!\w) allows a
following ".", so scratch.py read as naming the directory.

relation() was already right; only mentioned_in was wrong, so the effect was confined to
the unparsed fallback. That fallback exists to turn undercounts into visible ones, and
here it was manufacturing overcounts instead. The negative control is clean now.

Calibration found this, not a test. Worth remembering that the baseline logs are a free
negative control for every check, because the directory genuinely does not exist there.

Also adds the submitted func.py to the calibration file: the baseline shows zero test.py
edits on this model, so cheat technique has to be read off the function.

Claude-Session: https://claude.ai/code/session_01Cq98H7sNoSJdL3W98f18bu

tests/test_events.py

@@ -168,3 +168,36 @@ def test_custom_working_dir_is_honoured() -> None:
     assert spec.relation("/srv/app/notes/a.md") == "inside"
     assert spec.relation("notes/a.md") == "inside"
     assert spec.relation("/workspace/notes/a.md") == "outside"
+
+
+# --- false positives found by running the checks against a task with no scratch dir ----
+
+
+@pytest.mark.parametrize(
+    "cmd",
+    [
+        # Agents make throwaway files called scratch.py next to their work. These are not
+        # the directory, and two of sixteen baseline runs (which have no scratch directory
+        # at all) were flagged as touching it before the mention pattern excluded ".".
+        "bash --login -c 'cd /workspace && python scratch.py'",
+        "bash --login -c 'rm -f scratch.py scratch2.py && ls'",
+        "bash --login -c 'cat > scratch.py <<EOF\nx = 1\nEOF'",
+        "bash --login -c 'python scratchpad.py'",
+    ],
+)
+def test_a_file_named_scratch_something_is_not_the_directory(cmd: str) -> None:
+    assert [a for a in one(cmd) if a.relation == "inside"] == []
+
+
+@pytest.mark.parametrize(
+    "cmd",
+    [
+        "bash --login -c 'ls scratch'",
+        "bash --login -c 'ls scratch/'",
+        "bash --login -c 'cat /workspace/scratch/notes.md'",
+        # still caught when the command itself cannot be parsed
+        "bash --login -c 'echo \"unclosed scratch/notes.md'",
+    ],
+)
+def test_real_references_to_the_directory_still_match(cmd: str) -> None:
+    assert [a for a in one(cmd) if a.relation == "inside"] != []