diff --git a/.github/workflows/test.sh b/.github/workflows/test.sh
deleted file mode 100644
index 4fff2d1..0000000
--- a/.github/workflows/test.sh
+++ /dev/null
@@ -1,122 +0,0 @@
-set -e
-
-PLATFORM=$(python -c 'import sys; print(sys.platform)')
-
-RED="\033[31;1m"
-BLUE="\033[34m"
-MAGENTA="\033[35m"
-NORMAL="\033[0m"
-
-ptest() { echo -e "\n${MAGENTA}[TEST] $@${NORMAL}\n" ; }
-perror() { echo -e "\n${RED}[ERROR] $@${NORMAL}\n" ; }
-prun() { echo -e "${BLUE}\$ $@ ${NORMAL}" ; "$@" ; }
-
-prun cd example_pkg
-
-ptest version command runs
-prun spin --version
-
-ptest build command runs
-pip install meson-python ninja
-
-
-# Test spin build + debug builds
-echo "Creating debug builds"
-prun spin build --gcov
-ptest Did the build folder get generated?
-if [ ! -d "build" ] || [ ! -d "build-install" ]; then
-    perror build and/or build-install folders did not get generated
-    exit 1
-else
-    echo "Yes"
-fi
-ptest Does the debug build contain gcov files?
-matching_files=$(find . -type f -name "*.gc*")
-if [ -z "$matching_files" ]; then
-    perror Debug files did not get generated
-    exit 1
-else
-    echo "Yes"
-fi
-
-ptest Does spin expand \$PYTHONPATH?
-SPIN_PYTHONPATH=$(spin run 'echo $PYTHONPATH')
-echo spin sees PYTHONPATH=\"${SPIN_PYTHONPATH}\"
-if [[ ${SPIN_PYTHONPATH} == "\$PYTHONPATH" ]]; then
-    perror Expected Python path, but got $SPIN_PYTHONPATH instead
-    exit 1
-fi
-
-ptest Does \$PYTHONPATH contains site-packages?
-if [[ ${SPIN_PYTHONPATH} == *"site-packages" ]]; then
-    echo "Yes"
-else
-    echo "No; it is $SPIN_PYTHONPATH"
-fi
-
-ptest Does \`spin run\` redirect only command output to stdout?
-# Once we're on Python >3.11, can replace syspath manipulation below with -P flag to Python
-VERSION=$(spin run python -c 'import sys; del sys.path[0]; import example_pkg; print(example_pkg.__version__)')
-if [[ $VERSION == "0.0.0dev0" ]]; then
-    echo "Yes"
-else
-    perror No, output is $VERSION
-    exit 1
-fi
-
-ptest Does spin detect conflict with editable install?
-prun pip install --quiet -e .
-OUT=$(spin run ls)
-if [[ $OUT == *"Warning! An editable installation"* ]]; then
-    echo "Yes"
-else
-    perror No
-    exit 1
-fi
-prun pip uninstall --quiet -y example_pkg
-
-if [[ $PLATFORM == linux || $PLATFORM == darwin ]]; then
-    # Detecting whether a file is executable is not that easy on Windows,
-    # as it seems to take into consideration whether that file is associated as an executable.
-    ptest Does \`spin run foo.py\` warn that \`spin run python foo.py\` is likely intended?
-    OUT=$( touch __foo.py && spin run __foo.py || true )
-    rm __foo.py
-    if [[ $OUT == *"Did you mean to call"* ]]; then
-        echo "Yes"
-    else
-        perror No, output is: $OUT
-        exit 1
-    fi
-fi
-
-ptest test command runs
-prun spin test
-
-ptest Does \`spin test\` work when PYTHONPATH is set?
-PYTHONPATH=./tmp spin test
-
-ptest sdist command runs
-prun spin sdist
-
-ptest example command runs
-prun spin example
-
-ptest docs command runs
-pip install --quiet sphinx
-prun spin docs
-
-ptest install command works
-prun spin install
-(cd /tmp ; [[ $(python -c 'import example_pkg; print(example_pkg.__version__)') == "0.0.0dev0" ]])
-prun pip uninstall -y --quiet example_pkg
-
-## Platform specialized tests
-
-if [[ $PLATFORM == linux ]]; then
-    ptest gdb command runs on linux
-    prun spin gdb -c 'import example_pkg; example_pkg.echo("hi")' -- --eval "run" --batch
-fi
-
-# if [[ $PLATFORM == darwin ]]; then
-
-# if [[ $PLATFORM =~ ^win.* ]]; then
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index aca2d77..8057750 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -29,6 +29,6 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install -y gdb
-      - name: Tests
+      - name: Tests PyTest
         run: |
-          pipx run nox --forcecolor -s tests
+          pipx run nox --forcecolor -s test
diff --git a/noxfile.py b/noxfile.py
index 51bbbbc..ea40094 100644
--- a/noxfile.py
+++ b/noxfile.py
@@ -2,10 +2,6 @@
 
 
 @nox.session
-def tests(session: nox.Session) -> None:
-    """
-    Run the unit and regular tests.
-    """
-    session.install(".", "pytest", "build")
+def test(session: nox.Session) -> None:
+    session.install(".", "pytest", "build", "meson-python", "ninja")
     session.run("pytest", "spin", *session.posargs)
-    session.run("bash", ".github/workflows/test.sh", external=True)
diff --git a/spin/tests/conftest.py b/spin/tests/conftest.py
new file mode 100644
index 0000000..ffafaad
--- /dev/null
+++ b/spin/tests/conftest.py
@@ -0,0 +1,20 @@
+import os
+
+import pytest
+
+from spin import util
+
+
+@pytest.fixture(autouse=True)
+def pre_post_test():
+    # Pre-test code
+    cwd = os.getcwd()
+    os.chdir("example_pkg")
+
+    try:
+        yield
+    finally:
+        # Post test code
+        os.chdir(cwd)
+        util.run(["git", "clean", "-xdf"], cwd="example_pkg")
+        os.chdir(cwd)
diff --git a/spin/tests/test_build_cmds.py b/spin/tests/test_build_cmds.py
new file mode 100644
index 0000000..f802618
--- /dev/null
+++ b/spin/tests/test_build_cmds.py
@@ -0,0 +1,159 @@
+import os
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+
+import pytest
+
+import spin as libspin
+from spin.cmds.util import run
+
+skip_on_windows = pytest.mark.skipif(
+    sys.platform.startswith("win"), reason="Skipped; platform is Windows"
+)
+
+on_linux = pytest.mark.skipif(
+    not sys.platform.startswith("linux"), reason="Skipped; platform not Linux"
+)
+
+
+def spin(*args, **user_kwargs):
+    default_kwargs = {
+        "stdout": subprocess.PIPE,
+        "stderr": subprocess.PIPE,
+        "sys_exit": True,
+    }
+    return run(["spin"] + list(args), **{**default_kwargs, **user_kwargs})
+
+
+def stdout(p):
+    return p.stdout.decode("utf-8").strip()
+
+
+def stderr(p):
+    return p.stderr.decode("utf-8").strip()
+
+
+def test_get_version():
+    p = spin("--version")
+    assert stdout(p) == f"spin {libspin.__version__}"
+
+
+def test_basic_build():
+    """Does the package build?"""
+    spin("build")
+
+    assert Path("build").exists(), "`build` folder not created after `spin build`"
+    assert Path(
+        "build-install"
+    ).exists(), "`build-install` folder not created after `spin build`"
+
+
+def test_debug_builds():
+    """Does spin generate gcov debug output files?"""
+    spin("build", "--gcov")
+
+    debug_files = Path(".").rglob("*.gcno")
+    assert len(list(debug_files)) != 0, "debug files not generated for gcov build"
+
+
+def test_expand_pythonpath():
+    """Does an $ENV_VAR get expanded in `spin run`?"""
+    output = spin("run", "echo $PYTHONPATH")
+    assert any(
+        p in stdout(output) for p in ("site-packages", "dist-packages")
+    ), f"Expected value of $PYTHONPATH, got {stdout(output)} instead"
+
+
+def test_run_stdout():
+    """Ensure `spin run` only includes command output on stdout."""
+    p = spin(
+        "run",
+        sys.executable,
+        "-c",
+        "import sys; del sys.path[0]; import example_pkg; print(example_pkg.__version__)",
+    )
+    assert (
+        stdout(p) == "0.0.0dev0"
+    ), f"`spin run` stdout did not yield version, but {stdout(p)}"
+
+
+def test_editable_conflict():
+    """Do we warn when a conflicting editable install is present?"""
+    try:
+        run(["pip", "install", "--quiet", "-e", "."])
+        assert "Warning! An editable installation" in stdout(
+            spin("run", "ls")
+        ), "Failed to detect and warn about editable install"
+    finally:
+        run(["pip", "uninstall", "--quiet", "-y", "example_pkg"])
+
+
+# Detecting whether a file is executable is not that easy on Windows,
+# as it seems to take into consideration whether that file is associated as an executable.
+@skip_on_windows
+def test_recommend_run_python():
+    """If `spin run file.py` is called, is `spin run python file.py` recommended?"""
+    with tempfile.NamedTemporaryFile(suffix=".py") as f:
+        p = spin("run", f.name, sys_exit=False)
+        assert "Did you mean to call" in stdout(
+            p
+        ), "Failed to recommend `python run python file.py`"
+
+
+def test_test():
+    """Does the test command run?"""
+    spin("test")
+
+
+def test_test_with_pythonpath():
+    """Does `spin test` work when PYTHONPATH is set?"""
+    spin("test", env={**os.environ, "PYTHONPATH": "/tmp"})
+
+
+def test_sdist():
+    spin("sdist")
+
+
+def test_example():
+    spin("example")
+
+
+def test_docs():
+    run(["pip", "install", "--quiet", "sphinx"])
+    spin("docs")
+
+
+def test_spin_install():
+    cwd = os.getcwd()
+    spin("install")
+    with tempfile.TemporaryDirectory() as d:
+        try:
+            os.chdir(d)
+            p = run(
+                [
+                    sys.executable,
+                    "-c",
+                    "import example_pkg; print(example_pkg.__version__)",
+                ],
+                stdout=subprocess.PIPE,
+            )
+            assert stdout(p) == "0.0.0dev0"
+        finally:
+            os.chdir(cwd)
+            run(["pip", "uninstall", "-y", "--quiet", "example_pkg"])
+
+
+@on_linux
+def test_gdb():
+    p = spin(
+        "gdb",
+        "-c",
+        'import example_pkg; example_pkg.echo("hi")',
+        "--",
+        "--eval",
+        "run",
+        "--batch",
+    )
+    assert "hi" in stdout(p)
diff --git a/spin/tests/util.py b/spin/tests/util.py
new file mode 100644
index 0000000..09a3d63
--- /dev/null
+++ b/spin/tests/util.py
@@ -0,0 +1,28 @@
+import os
+import subprocess
+
+from spin.cmds import util
+
+PKG_NAME = "example_pkg"
+
+
+def assert_cmd(cmd, *args, **kwargs) -> str:
+    cwd = os.getcwd()
+    p = util.run(
+        cmd,
+        *args,
+        cwd=PKG_NAME,
+        replace=False,
+        sys_exit=False,
+        output=False,
+        echo=True,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+        **kwargs,
+    )
+    os.chdir(cwd)
+    stdout = p.stdout.decode("utf-8").strip()
+    if not p.returncode == 0:
+        print(stdout)
+        raise AssertionError(f"[{cmd}] failed with exit code {p.returncode}")
+    return p