feat(plugins): convert disk-guardian skill into a bundled plugin

Rewires @LVT382009's disk-guardian (PR #12212) from a skill-plus-script
into a plugin that runs entirely via hooks — no agent compliance needed.

- post_tool_call hook auto-tracks files created by write_file / terminal
  / patch when they match test_/tmp_/*.test.* patterns under HERMES_HOME
- on_session_end hook runs cmd_quick cleanup when test files were
  auto-tracked during the turn; stays quiet otherwise
- /disk-guardian slash command keeps status / dry-run / quick / deep /
  track / forget for manual use
- Deterministic cleanup rules, path safety, atomic writes, and audit
  logging preserved from the original contribution
- Protect well-known top-level state dirs (logs/, memories/, sessions/,
  cron/, cache/, etc.) from empty-dir removal so fresh installs don't
  get gutted on first session end

The plugin system gains a bundled-plugin discovery path (<repo>/plugins/
<name>/) alongside user/project/entry-point sources. Memory and
context_engine subdirs are skipped — they keep their own discovery
paths. HERMES_DISABLE_BUNDLED_PLUGINS=1 suppresses the scan; the test
conftest sets it by default so existing plugin tests stay clean.

Co-authored-by: LVT382009 <levantam.98.2324@gmail.com>

tests/plugins/test_disk_guardian_plugin.py

@@ -0,0 +1,426 @@
+"""Tests for the disk-guardian plugin.
+
+Covers the bundled plugin at ``plugins/disk-guardian/``:
+
+  * ``disk_guardian`` library: track / forget / dry_run / quick / status,
+    ``is_safe_path`` and ``guess_category`` filtering.
+  * Plugin ``__init__``: ``post_tool_call`` hook auto-tracks files created
+    by ``write_file`` / ``terminal``; ``on_session_end`` hook runs quick
+    cleanup when anything was tracked during the turn.
+  * Slash command handler: status / dry-run / quick / track / forget /
+    unknown subcommand behaviours.
+  * Bundled-plugin discovery via ``PluginManager.discover_and_load``.
+"""
+
+import importlib
+import json
+import os
+import sys
+from pathlib import Path
+
+import pytest
+
+
+@pytest.fixture(autouse=True)
+def _isolate_env(tmp_path, monkeypatch):
+    """Isolate HERMES_HOME + clear plugin module cache for each test."""
+    hermes_home = tmp_path / ".hermes"
+    hermes_home.mkdir()
+    monkeypatch.setenv("HERMES_HOME", str(hermes_home))
+    # Drop the disk-guardian modules so each test re-imports fresh.
+    for mod in list(sys.modules.keys()):
+        if mod.startswith("hermes_plugins.disk_guardian") or mod == "plugins.disk_guardian":
+            del sys.modules[mod]
+    yield hermes_home
+
+
+def _load_lib():
+    """Import the plugin's library module directly from the repo path."""
+    repo_root = Path(__file__).resolve().parents[2]
+    lib_path = repo_root / "plugins" / "disk-guardian" / "disk_guardian.py"
+    spec = importlib.util.spec_from_file_location(
+        "disk_guardian_under_test", lib_path
+    )
+    mod = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(mod)
+    return mod
+
+
+def _load_plugin_init():
+    """Import the plugin's __init__.py (which depends on the library)."""
+    repo_root = Path(__file__).resolve().parents[2]
+    plugin_dir = repo_root / "plugins" / "disk-guardian"
+    # Use the PluginManager's module naming convention so relative imports work.
+    spec = importlib.util.spec_from_file_location(
+        "hermes_plugins.disk_guardian",
+        plugin_dir / "__init__.py",
+        submodule_search_locations=[str(plugin_dir)],
+    )
+    # Ensure parent namespace package exists for the relative `. import disk_guardian`
+    import types
+    if "hermes_plugins" not in sys.modules:
+        ns = types.ModuleType("hermes_plugins")
+        ns.__path__ = []
+        sys.modules["hermes_plugins"] = ns
+    mod = importlib.util.module_from_spec(spec)
+    mod.__package__ = "hermes_plugins.disk_guardian"
+    mod.__path__ = [str(plugin_dir)]
+    sys.modules["hermes_plugins.disk_guardian"] = mod
+    spec.loader.exec_module(mod)
+    return mod
+
+
+# ---------------------------------------------------------------------------
+# Library tests
+# ---------------------------------------------------------------------------
+
+class TestIsSafePath:
+    def test_accepts_path_under_hermes_home(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "subdir" / "file.txt"
+        p.parent.mkdir()
+        p.write_text("x")
+        assert dg.is_safe_path(p) is True
+
+    def test_rejects_outside_hermes_home(self, _isolate_env):
+        dg = _load_lib()
+        assert dg.is_safe_path(Path("/etc/passwd")) is False
+
+    def test_accepts_tmp_hermes_prefix(self, _isolate_env, tmp_path):
+        dg = _load_lib()
+        assert dg.is_safe_path(Path("/tmp/hermes-abc/x.log")) is True
+
+    def test_rejects_plain_tmp(self, _isolate_env):
+        dg = _load_lib()
+        assert dg.is_safe_path(Path("/tmp/other.log")) is False
+
+    def test_rejects_windows_mount(self, _isolate_env):
+        dg = _load_lib()
+        assert dg.is_safe_path(Path("/mnt/c/Users/x/test.txt")) is False
+
+
+class TestGuessCategory:
+    def test_test_prefix(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "test_foo.py"
+        p.write_text("x")
+        assert dg.guess_category(p) == "test"
+
+    def test_tmp_prefix(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "tmp_foo.log"
+        p.write_text("x")
+        assert dg.guess_category(p) == "test"
+
+    def test_dot_test_suffix(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "mything.test.js"
+        p.write_text("x")
+        assert dg.guess_category(p) == "test"
+
+    def test_skips_protected_top_level(self, _isolate_env):
+        dg = _load_lib()
+        logs_dir = _isolate_env / "logs"
+        logs_dir.mkdir()
+        p = logs_dir / "test_log.txt"
+        p.write_text("x")
+        # Even though it matches test_* pattern, logs/ is excluded.
+        assert dg.guess_category(p) is None
+
+    def test_cron_subtree_categorised(self, _isolate_env):
+        dg = _load_lib()
+        cron_dir = _isolate_env / "cron"
+        cron_dir.mkdir()
+        p = cron_dir / "job_output.md"
+        p.write_text("x")
+        assert dg.guess_category(p) == "cron-output"
+
+    def test_ordinary_file_returns_none(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "notes.md"
+        p.write_text("x")
+        assert dg.guess_category(p) is None
+
+
+class TestTrackForgetQuick:
+    def test_track_then_quick_deletes_test(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "test_a.py"
+        p.write_text("x")
+        assert dg.track(str(p), "test", silent=True) is True
+        summary = dg.quick()
+        assert summary["deleted"] == 1
+        assert not p.exists()
+
+    def test_track_dedup(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "test_a.py"
+        p.write_text("x")
+        assert dg.track(str(p), "test", silent=True) is True
+        # Second call returns False (already tracked)
+        assert dg.track(str(p), "test", silent=True) is False
+
+    def test_track_rejects_outside_home(self, _isolate_env):
+        dg = _load_lib()
+        # /etc/hostname exists on most Linux boxes; fall back if not.
+        outside = "/etc/hostname" if Path("/etc/hostname").exists() else "/etc/passwd"
+        assert dg.track(outside, "test", silent=True) is False
+
+    def test_track_skips_missing(self, _isolate_env):
+        dg = _load_lib()
+        assert dg.track(str(_isolate_env / "nope.txt"), "test", silent=True) is False
+
+    def test_forget_removes_entry(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "keep.tmp"
+        p.write_text("x")
+        dg.track(str(p), "temp", silent=True)
+        assert dg.forget(str(p)) == 1
+        assert p.exists()  # forget does NOT delete the file
+
+    def test_quick_preserves_unexpired_temp(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "fresh.tmp"
+        p.write_text("x")
+        dg.track(str(p), "temp", silent=True)
+        summary = dg.quick()
+        assert summary["deleted"] == 0
+        assert p.exists()
+
+    def test_quick_preserves_protected_top_level_dirs(self, _isolate_env):
+        dg = _load_lib()
+        for d in ("logs", "memories", "sessions", "cron", "cache"):
+            (_isolate_env / d).mkdir()
+        dg.quick()
+        for d in ("logs", "memories", "sessions", "cron", "cache"):
+            assert (_isolate_env / d).exists(), f"{d}/ should be preserved"
+
+
+class TestStatus:
+    def test_empty_status(self, _isolate_env):
+        dg = _load_lib()
+        s = dg.status()
+        assert s["total_tracked"] == 0
+        assert s["top10"] == []
+
+    def test_status_with_entries(self, _isolate_env):
+        dg = _load_lib()
+        p = _isolate_env / "big.tmp"
+        p.write_text("y" * 100)
+        dg.track(str(p), "temp", silent=True)
+        s = dg.status()
+        assert s["total_tracked"] == 1
+        assert len(s["top10"]) == 1
+        rendered = dg.format_status(s)
+        assert "temp" in rendered
+        assert "big.tmp" in rendered
+
+
+class TestDryRun:
+    def test_classifies_by_category(self, _isolate_env):
+        dg = _load_lib()
+        test_f = _isolate_env / "test_x.py"
+        test_f.write_text("x")
+        big = _isolate_env / "big.bin"
+        big.write_bytes(b"z" * 10)
+        dg.track(str(test_f), "test", silent=True)
+        dg.track(str(big), "other", silent=True)
+        auto, prompt = dg.dry_run()
+        # test → auto, other → neither (doesn't hit any rule)
+        assert any(i["path"] == str(test_f) for i in auto)
+
+
+# ---------------------------------------------------------------------------
+# Plugin hooks tests
+# ---------------------------------------------------------------------------
+
+class TestPostToolCallHook:
+    def test_write_file_test_pattern_tracked(self, _isolate_env):
+        pi = _load_plugin_init()
+        p = _isolate_env / "test_created.py"
+        p.write_text("x")
+        pi._on_post_tool_call(
+            tool_name="write_file",
+            args={"path": str(p), "content": "x"},
+            result="OK",
+            task_id="t1", session_id="s1",
+        )
+        tracked_file = _isolate_env / "disk-guardian" / "tracked.json"
+        data = json.loads(tracked_file.read_text())
+        assert len(data) == 1
+        assert data[0]["category"] == "test"
+
+    def test_write_file_non_test_not_tracked(self, _isolate_env):
+        pi = _load_plugin_init()
+        p = _isolate_env / "notes.md"
+        p.write_text("x")
+        pi._on_post_tool_call(
+            tool_name="write_file",
+            args={"path": str(p), "content": "x"},
+            result="OK",
+            task_id="t2", session_id="s2",
+        )
+        tracked_file = _isolate_env / "disk-guardian" / "tracked.json"
+        assert not tracked_file.exists() or tracked_file.read_text().strip() == "[]"
+
+    def test_terminal_command_picks_up_paths(self, _isolate_env):
+        pi = _load_plugin_init()
+        p = _isolate_env / "tmp_created.log"
+        p.write_text("x")
+        pi._on_post_tool_call(
+            tool_name="terminal",
+            args={"command": f"touch {p}"},
+            result=f"created {p}\n",
+            task_id="t3", session_id="s3",
+        )
+        tracked_file = _isolate_env / "disk-guardian" / "tracked.json"
+        data = json.loads(tracked_file.read_text())
+        assert any(Path(i["path"]) == p.resolve() for i in data)
+
+    def test_ignores_unrelated_tool(self, _isolate_env):
+        pi = _load_plugin_init()
+        pi._on_post_tool_call(
+            tool_name="read_file",
+            args={"path": str(_isolate_env / "test_x.py")},
+            result="contents",
+            task_id="t4", session_id="s4",
+        )
+        # read_file should never trigger tracking.
+        tracked_file = _isolate_env / "disk-guardian" / "tracked.json"
+        assert not tracked_file.exists() or tracked_file.read_text().strip() == "[]"
+
+
+class TestOnSessionEndHook:
+    def test_runs_quick_when_test_files_tracked(self, _isolate_env):
+        pi = _load_plugin_init()
+        p = _isolate_env / "test_cleanup.py"
+        p.write_text("x")
+        pi._on_post_tool_call(
+            tool_name="write_file",
+            args={"path": str(p), "content": "x"},
+            result="OK",
+            task_id="", session_id="s1",
+        )
+        assert p.exists()
+        pi._on_session_end(session_id="s1", completed=True, interrupted=False)
+        assert not p.exists(), "test file should be auto-deleted"
+
+    def test_noop_when_no_test_tracked(self, _isolate_env):
+        pi = _load_plugin_init()
+        # Nothing tracked → on_session_end should not raise.
+        pi._on_session_end(session_id="empty", completed=True, interrupted=False)
+
+
+# ---------------------------------------------------------------------------
+# Slash command
+# ---------------------------------------------------------------------------
+
+class TestSlashCommand:
+    def test_help(self, _isolate_env):
+        pi = _load_plugin_init()
+        out = pi._handle_slash("help")
+        assert "disk-guardian" in out
+        assert "status" in out
+
+    def test_status_empty(self, _isolate_env):
+        pi = _load_plugin_init()
+        out = pi._handle_slash("status")
+        assert "nothing tracked" in out
+
+    def test_track_rejects_missing(self, _isolate_env):
+        pi = _load_plugin_init()
+        out = pi._handle_slash(
+            f"track {_isolate_env / 'nope.txt'} temp"
+        )
+        assert "Not tracked" in out
+
+    def test_track_rejects_bad_category(self, _isolate_env):
+        pi = _load_plugin_init()
+        p = _isolate_env / "a.tmp"
+        p.write_text("x")
+        out = pi._handle_slash(f"track {p} banana")
+        assert "Unknown category" in out
+
+    def test_track_and_forget(self, _isolate_env):
+        pi = _load_plugin_init()
+        p = _isolate_env / "a.tmp"
+        p.write_text("x")
+        out = pi._handle_slash(f"track {p} temp")
+        assert "Tracked" in out
+        out = pi._handle_slash(f"forget {p}")
+        assert "Removed 1" in out
+
+    def test_unknown_subcommand(self, _isolate_env):
+        pi = _load_plugin_init()
+        out = pi._handle_slash("foobar")
+        assert "Unknown subcommand" in out
+
+    def test_quick_on_empty(self, _isolate_env):
+        pi = _load_plugin_init()
+        out = pi._handle_slash("quick")
+        assert "Cleaned 0 files" in out
+
+
+# ---------------------------------------------------------------------------
+# Bundled-plugin discovery
+# ---------------------------------------------------------------------------
+
+class TestBundledDiscovery:
+    def test_disk_guardian_is_discovered_as_bundled(self, _isolate_env, monkeypatch):
+        # The default hermetic conftest disables bundled plugin discovery.
+        # This test specifically exercises it, so clear the suppression.
+        monkeypatch.delenv("HERMES_DISABLE_BUNDLED_PLUGINS", raising=False)
+        # Reset plugin manager state so discovery runs fresh.
+        for mod in list(sys.modules.keys()):
+            if mod.startswith("hermes_cli.plugins") or mod == "plugins":
+                del sys.modules[mod]
+
+        repo_root = Path(__file__).resolve().parents[2]
+        sys.path.insert(0, str(repo_root))
+        try:
+            from hermes_cli import plugins as pmod
+            mgr = pmod.PluginManager()
+            mgr.discover_and_load()
+            assert "disk-guardian" in mgr._plugins
+            loaded = mgr._plugins["disk-guardian"]
+            assert loaded.manifest.source == "bundled"
+            assert loaded.enabled
+            assert "post_tool_call" in loaded.hooks_registered
+            assert "on_session_end" in loaded.hooks_registered
+            assert "disk-guardian" in loaded.commands_registered
+        finally:
+            sys.path.pop(0)
+
+    def test_memory_and_context_engine_subdirs_skipped(self, _isolate_env, monkeypatch):
+        """Bundled scan must NOT pick up plugins/memory or plugins/context_engine
+        as top-level plugins — they have their own discovery paths."""
+        monkeypatch.delenv("HERMES_DISABLE_BUNDLED_PLUGINS", raising=False)
+        for mod in list(sys.modules.keys()):
+            if mod.startswith("hermes_cli.plugins") or mod == "plugins":
+                del sys.modules[mod]
+        repo_root = Path(__file__).resolve().parents[2]
+        sys.path.insert(0, str(repo_root))
+        try:
+            from hermes_cli import plugins as pmod
+            mgr = pmod.PluginManager()
+            mgr.discover_and_load()
+            assert "memory" not in mgr._plugins
+            assert "context_engine" not in mgr._plugins
+        finally:
+            sys.path.pop(0)
+
+    def test_bundled_scan_suppressed_by_env_var(self, _isolate_env, monkeypatch):
+        """HERMES_DISABLE_BUNDLED_PLUGINS=1 suppresses bundled discovery."""
+        monkeypatch.setenv("HERMES_DISABLE_BUNDLED_PLUGINS", "1")
+        for mod in list(sys.modules.keys()):
+            if mod.startswith("hermes_cli.plugins") or mod == "plugins":
+                del sys.modules[mod]
+        repo_root = Path(__file__).resolve().parents[2]
+        sys.path.insert(0, str(repo_root))
+        try:
+            from hermes_cli import plugins as pmod
+            mgr = pmod.PluginManager()
+            mgr.discover_and_load()
+            assert "disk-guardian" not in mgr._plugins
+        finally:
+            sys.path.pop(0)