fix(update): identify source archives and exclude dev metadata

- Embed the archived commit SHA to prevent false first-launch updates
- Exclude editor and agent directories from source and update archives
- Add coverage for archive metadata and installer filtering

.agents/skills/shipped-data-assets/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: shipped-data-assets
+description: Keep shipped data assets present in git branch archives and tool updates. Use when adding or changing defaults under data/, editing their .gitignore rules, changing the update installer, or updating bundled-data coverage in tests/test_bundled_updates.py.
+---
+
+# Shipped Data Assets
+
+Tool auto-update downloads a git branch archive. Files matched by `.gitignore` are omitted from that archive and never reach users.
+
+## Maintain shipped defaults
+
+- Add a `.gitignore` negation for every shipped path under `data/`, following existing patterns such as:
+  - `!data/skills/**`
+  - `!data/help/**`
+  - `!data/vocab_base.txt`
+  - `!data/translation_contexts.json`
+- Do not rely on `UpdateThread.should_install` alone; ignored files never appear in the downloaded archive.
+- Extend `_SHIPPED_DATA_FILES` or `ShippedDataTrackingTests` in `tests/test_bundled_updates.py` when adding a shipped path.
+- Keep user-local files ignored and protected, including `data/vocab.txt`, `data/last_update_sha.txt`, `data/wolf_*.json`, and `data/api_keys.json`.
+
+## Verify tracking
+
+Run `git check-ignore -- <shipped-path>` for each affected file. A shipped path must exit with status 1, indicating that it is not ignored.
+
+Run the bundled-update tests after changing shipped paths, ignore rules, or installer behavior:
+
+```bash
+python -m unittest tests.test_bundled_updates.ShippedDataTrackingTests
+```

.cursor/rules/shipped-data-assets.mdc

@@ -1,27 +0,0 @@
----
-description: Shipped data/ assets must be git-trackable so tool updates include them
-globs: "{.gitignore,data/**,gui/guide_tab.py,gui/main.py,tests/test_bundled_updates.py}"
-alwaysApply: false
----
-
-# Shipped data/ and tool updates
-
-Tool auto-update downloads a **git branch archive**.
-Anything matched by `.gitignore` is omitted from that archive and never reaches users.
-
-## When adding or changing shipped defaults under `data/`
-
-- Add a negation under `.gitignore` (same pattern as skills/help):
-  - `!data/skills/**`
-  - `!data/help/**`
-  - `!data/vocab_base.txt`
-  - `!data/translation_contexts.json`
-- Do **not** rely on `UpdateThread.should_install` alone - ignored files never appear in the zip.
-- Extend `_SHIPPED_DATA_FILES` / `ShippedDataTrackingTests` in `tests/test_bundled_updates.py` when adding a new shipped path.
-- User-local files stay ignored/protected: `data/vocab.txt`, `data/last_update_sha.txt`, `data/wolf_*.json`, `data/api_keys.json`.
-
-## Quick check
-
-```bash
-git check-ignore -- data/help/index.json   # must exit 1 (not ignored)
-```

.git_archival.txt

@@ -0,0 +1 @@
+node: $Format:%H$

.gitattributes

@@ -1,2 +1,14 @@
 # Cmd.exe parses LF-only .bat files incorrectly on Windows (joined tokens / mangled commands).
 *.bat text eol=crlf
+
+# Keep development-only agent/editor metadata out of user-facing source and
+# update archives.
+.agents export-ignore
+.codex export-ignore
+.cursor export-ignore
+.github export-ignore
+.vscode export-ignore
+
+# Expand the exact archived commit so a fresh source download can identify the
+# version it already contains without committing a self-referential literal SHA.
+.git_archival.txt export-subst

.gitignore

@@ -19,6 +19,7 @@ __pycache__
 !assets/*.png
 !assets/engine_icons/**
 !.pre-commit-config.yaml
+!.git_archival.txt
 !data/skills/**
 !data/help/**
 !data/vocab_base.txt

gui/main.py

@@ -53,8 +53,7 @@ def check_tool_update() -> str | None:
     """Return latest commit SHA when a tool update is available, else None."""
     try:
         latest_sha = UpdateThread.fetch_latest_sha()
-        sha_path = Path(UpdateThread.SHA_FILE)
+        current_sha = UpdateThread.read_installed_sha()
-        current_sha = sha_path.read_text().strip() if sha_path.is_file() else ""
         if latest_sha != current_sha:
             return latest_sha
     except Exception:
@@ -82,9 +81,21 @@ class UpdateThread(QThread):
     # "dazedtl"). Resolve dynamically so a rename cannot silently no-op again.
     ARCHIVE_ROOT = "dazedtl"
     SHA_FILE = str(LAST_UPDATE_SHA_PATH)
+    ARCHIVE_SHA_FILE = str(PROJECT_ROOT / ".git_archival.txt")
 
     # Top-level paths that should never be touched during update
-    PROTECTED_TOP = {".env", "venv", "log", "files", "translated"}
+    PROTECTED_TOP = {
+        ".agents",
+        ".codex",
+        ".cursor",
+        ".env",
+        ".github",
+        ".vscode",
+        "venv",
+        "log",
+        "files",
+        "translated",
+    }
 
     # User-local files under data/ that must not be overwritten.
     # Shipped defaults (translation_contexts.json, skills/*.md, help/*, vocab_base.txt, …)
@@ -185,14 +196,45 @@ class UpdateThread(QThread):
         with urllib.request.urlopen(req, timeout=15) as resp:
             return json.loads(resp.read())["commit"]["id"]
 
+    @staticmethod
+    def _read_archive_sha(path: Path) -> str:
+        """Read an expanded git-archive commit ID, or return an empty string."""
+        if not path.is_file():
+            return ""
+        try:
+            for line in path.read_text(encoding="utf-8").splitlines():
+                key, separator, value = line.partition(":")
+                candidate = value.strip()
+                if (
+                    separator
+                    and key.strip() == "node"
+                    and re.fullmatch(r"[0-9a-fA-F]{40,64}", candidate)
+                ):
+                    return candidate
+        except OSError:
+            pass
+        return ""
+
+    @classmethod
+    def read_installed_sha(cls) -> str:
+        """Return the updated SHA, falling back to source-archive metadata."""
+        runtime_path = Path(cls.SHA_FILE)
+        if runtime_path.is_file():
+            try:
+                runtime_sha = runtime_path.read_text(encoding="utf-8").strip()
+                if runtime_sha:
+                    return runtime_sha
+            except OSError:
+                pass
+        return cls._read_archive_sha(Path(cls.ARCHIVE_SHA_FILE))
+
     # ------------------------------------------------------------------ #
 
     def _fetch_latest_sha(self):
         return self.fetch_latest_sha()
 
     def _read_stored_sha(self):
-        p = Path(self.SHA_FILE)
+        return self.read_installed_sha()
-        return p.read_text().strip() if p.exists() else ""
 
     def _download_archive(self, zip_path: Path):
         req = urllib.request.Request(
@@ -493,10 +535,7 @@ class UpdateDialog(QDialog):
 
     @classmethod
     def _installed_sha_display(cls) -> str:
-        sha_path = Path(UpdateThread.SHA_FILE)
+        return cls._short_sha(UpdateThread.read_installed_sha())
-        if sha_path.is_file():
-            return cls._short_sha(sha_path.read_text().strip())
-        return "Not recorded"
 
     def _set_stage_pills(self, active_stage: str | None = None):
         stage_order = list(UpdateThread.STAGES)

gui/translation_tab.py

@@ -19,6 +19,7 @@ import traceback
 import signal
 import multiprocessing
 import re
+from importlib import import_module
 from colorama import Fore
 from tqdm import tqdm
 from dotenv import load_dotenv
@@ -75,6 +76,39 @@ BATCH_COLLECT_LIVE_CHARGE_NOTE = (
 )
 
 
+TRANSLATION_MODULE_SPECS = (
+    ("RPG Maker MV/MZ", (".json",), "modules.rpgmakermvmz", "handleMVMZ"),
+    ("CSV", (".csv",), "modules.csv", "handleCSV"),
+    ("Tyrano", (".ks",), "modules.tyrano", "handleTyrano"),
+    ("Kirikiri", (".ks",), "modules.kirikiri", "handleKirikiri"),
+    ("JSON", (".json",), "modules.json", "handleJSON"),
+    ("Lune", (".l",), "modules.lune", "handleLune"),
+    ("Yuris", (".json",), "modules.yuris", "handleYuris"),
+    ("NScript", (".nscript",), "modules.nscript", "handleOnscripter"),
+    ("Wolf RPG (WolfDawn)", (".json",), "modules.wolfdawn", "handleWolfDawn"),
+    ("Wolf RPG", (".json",), "modules.wolf", "handleWOLF"),
+    ("Wolf RPG 2", (".txt",), "modules.wolf2", "handleWOLF2"),
+    ("Regex", (".txt", ".json", ".script", ".csv"), "modules.regex", "handleRegex"),
+    ("Text", (".txt", ".srt"), "modules.text", "handleText"),
+    ("RenPy", (".rpy",), "modules.renpy", "handleRenpy"),
+    ("Unity", (".unity",), "modules.unity", "handleUnity"),
+    ("Images", (".png", ".jpg", ".jpeg"), "modules.images", "handleImages"),
+    ("RPG Maker Plugin", (".js",), "modules.rpgmakerplugin", "handlePlugin"),
+    ("Aquedi4 Prepared JSON", (".json",), "modules.aquedi4", "handleAquedi4"),
+    ("SRPG Studio", (".json",), "modules.srpg", "handleSRPG"),
+)
+
+
+def _lazy_module_handler(module_name, handler_name):
+    """Return a handler that imports its engine only when translation starts."""
+
+    def run(*args, **kwargs):
+        module = import_module(module_name)
+        return getattr(module, handler_name)(*args, **kwargs)
+
+    return run
+
+
 class _ShimLabel:
     """Plain stand-in for QLabel used by Files-tab row helpers (no real Qt widget)."""
 
@@ -1647,63 +1681,25 @@ class TranslationTab(QWidget):
         
     def setup_module_list(self):
         """Set up the module selection list."""
-        # Import modules to get the list
+        # Engine modules read translation settings during import. A downloaded
-        try:
+        # source archive intentionally has no .env yet, so importing every
-            sys.path.append(str(self.project_root))
+        # engine here used to abort discovery and leave only the fallback item.
-            from modules.rpgmakermvmz import handleMVMZ
+        # Keep the UI registry independent and defer imports until a handler is
-            from modules.csv import handleCSV
+        # actually used.
-            from modules.tyrano import handleTyrano
+        self.modules = [
-            from modules.kirikiri import handleKirikiri
+            [
-            from modules.json import handleJSON
+                display_name,
-            from modules.lune import handleLune
+                list(extensions),
-            from modules.yuris import handleYuris
+                _lazy_module_handler(module_name, handler_name),
-            from modules.nscript import handleOnscripter
-            from modules.wolf import handleWOLF
-            from modules.wolf2 import handleWOLF2
-            from modules.wolfdawn import handleWolfDawn
-            from modules.regex import handleRegex
-            from modules.text import handleText
-            from modules.renpy import handleRenpy
-            from modules.unity import handleUnity
-            from modules.images import handleImages
-            from modules.rpgmakerplugin import handlePlugin
-            from modules.aquedi4 import handleAquedi4
-            from modules.srpg import handleSRPG
-            
-            self.modules = [
-                ["RPG Maker MV/MZ", [".json"], handleMVMZ],
-                ["CSV", [".csv"], handleCSV],
-                ["Tyrano", [".ks"], handleTyrano],
-                ["Kirikiri", [".ks"], handleKirikiri],
-                ["JSON", [".json"], handleJSON],
-                ["Lune", [".l"], handleLune],
-                ["Yuris", [".json"], handleYuris],
-                ["NScript", [".nscript"], handleOnscripter],
-                ["Wolf RPG (WolfDawn)", [".json"], handleWolfDawn],
-                ["Wolf RPG", [".json"], handleWOLF],
-                ["Wolf RPG 2", [".txt"], handleWOLF2],
-                ["Regex", [".txt", ".json", ".script", ".csv"], handleRegex],
-                ["Text", [".txt", ".srt"], handleText],
-                ["RenPy", [".rpy"], handleRenpy],
-                ["Unity", [".unity"], handleUnity],
-                ["Images", [".png", ".jpg", ".jpeg"], handleImages],
-                ["RPG Maker Plugin", [".js"], handlePlugin],
-                ["Aquedi4 Prepared JSON", [".json"], handleAquedi4],
-                ["SRPG Studio", [".json"], handleSRPG],
             ]
-            
+            for display_name, extensions, module_name, handler_name
-            for module in self.modules:
+            in TRANSLATION_MODULE_SPECS
-                extensions = ", ".join(module[1])
+        ]
-                self.module_combo.addItem(f"{module[0]} ({extensions})")
+
-            if self.module_combo.count():
+        for module in self.modules:
-                self._on_module_changed(self.module_combo.currentText())
+            extensions = ", ".join(module[1])
-                
+            self.module_combo.addItem(f"{module[0]} ({extensions})")
-        except Exception as e:
+        if self.module_combo.count():
-            # Store error for later logging since log_display might not exist yet
-            self.module_load_error = f"Warning: Could not load modules: {str(e)}"
-            # Add a default option
-            self.module_combo.addItem("RPG Maker MV/MZ (.json)")
-            self.modules = [["RPG Maker MV/MZ", [".json"], None]]
             self._on_module_changed(self.module_combo.currentText())
 
     def _on_module_changed(self, text: str):

tests/test_bundled_updates.py

@@ -37,6 +37,7 @@ class UpdateThreadInstallFilterTests(unittest.TestCase):
 
         for rel in (
             *_SHIPPED_DATA_FILES,
+            ".git_archival.txt",
             "gui/main.py",
             "gameupdate/GameUpdate.bat",
             "gameupdate/gameupdate/patch.ps1",
@@ -58,6 +59,11 @@ class UpdateThreadInstallFilterTests(unittest.TestCase):
             "log/translations.txt",
             "files/Map001.json",
             "translated/Map001.json",
+            ".agents/skills/shipped-data-assets/SKILL.md",
+            ".codex/config.toml",
+            ".cursor/rules/project.mdc",
+            ".github/workflows/tests.yml",
+            ".vscode/settings.json",
         ):
             with self.subTest(rel=rel):
                 self.assertFalse(UpdateThread.should_install(Path(rel)))
@@ -180,8 +186,49 @@ class ShippedDataTrackingTests(unittest.TestCase):
                 )
 
 
+class SourceArchiveMetadataTests(unittest.TestCase):
+    """Source archives identify their commit and omit development metadata."""
+
+    def test_archival_info_is_configured_for_sha_expansion(self):
+        archival_info = _REPO_ROOT / ".git_archival.txt"
+        self.assertTrue(archival_info.is_file())
+        self.assertIn("$Format:%H$", archival_info.read_text(encoding="utf-8"))
+
+        ignored = subprocess.run(
+            ["git", "check-ignore", "--", ".git_archival.txt"],
+            cwd=_REPO_ROOT,
+            capture_output=True,
+            text=True,
+            check=False,
+        )
+        self.assertNotEqual(ignored.returncode, 0)
+
+        result = subprocess.run(
+            ["git", "check-attr", "export-subst", "--", ".git_archival.txt"],
+            cwd=_REPO_ROOT,
+            capture_output=True,
+            text=True,
+            check=False,
+        )
+        self.assertEqual(result.returncode, 0)
+        self.assertIn("export-subst: set", result.stdout)
+
+    def test_development_directories_are_export_ignored(self):
+        for rel in (".agents", ".codex", ".cursor", ".github", ".vscode"):
+            with self.subTest(rel=rel):
+                result = subprocess.run(
+                    ["git", "check-attr", "export-ignore", "--", rel],
+                    cwd=_REPO_ROOT,
+                    capture_output=True,
+                    text=True,
+                    check=False,
+                )
+                self.assertEqual(result.returncode, 0)
+                self.assertIn("export-ignore: set", result.stdout)
+
+
 class CheckToolUpdateTests(unittest.TestCase):
-    """Tool update check compares remote SHA to data/last_update_sha.txt only."""
+    """Tool update checks runtime state, then source-archive commit metadata."""
 
     def test_returns_sha_when_remote_differs(self):
         with tempfile.TemporaryDirectory() as raw:
@@ -207,6 +254,43 @@ class CheckToolUpdateTests(unittest.TestCase):
                 ):
                     self.assertIsNone(check_tool_update())
 
+    def test_archive_sha_prevents_false_update_on_first_launch(self):
+        archived_sha = "a" * 40
+        with tempfile.TemporaryDirectory() as raw:
+            base = Path(raw)
+            sha_file = base / "missing-last-update-sha.txt"
+            archive_file = base / ".git_archival.txt"
+            archive_file.write_text(f"node: {archived_sha}\n", encoding="utf-8")
+            from gui.main import UpdateThread, check_tool_update
+
+            with patch.multiple(
+                UpdateThread,
+                SHA_FILE=str(sha_file),
+                ARCHIVE_SHA_FILE=str(archive_file),
+            ):
+                with patch.object(
+                    UpdateThread, "fetch_latest_sha", return_value=archived_sha
+                ):
+                    self.assertIsNone(check_tool_update())
+
+    def test_unexpanded_archive_sha_is_ignored(self):
+        latest_sha = "b" * 40
+        with tempfile.TemporaryDirectory() as raw:
+            base = Path(raw)
+            archive_file = base / ".git_archival.txt"
+            archive_file.write_text("node: $Format:%H$\n", encoding="utf-8")
+            from gui.main import UpdateThread, check_tool_update
+
+            with patch.multiple(
+                UpdateThread,
+                SHA_FILE=str(base / "missing-last-update-sha.txt"),
+                ARCHIVE_SHA_FILE=str(archive_file),
+            ):
+                with patch.object(
+                    UpdateThread, "fetch_latest_sha", return_value=latest_sha
+                ):
+                    self.assertEqual(check_tool_update(), latest_sha)
+
     def test_returns_none_on_fetch_failure(self):
         from gui.main import UpdateThread, check_tool_update
 

tests/test_translation_engine_dropdown.py

@@ -0,0 +1,58 @@
+"""Regression tests for Translation-tab engine discovery."""
+
+from __future__ import annotations
+
+import os
+import unittest
+from unittest.mock import patch
+
+os.environ.setdefault("QT_QPA_PLATFORM", "offscreen")
+
+try:
+    from PyQt5.QtWidgets import QApplication
+
+    from gui.translation_tab import TRANSLATION_MODULE_SPECS, TranslationTab
+
+    _HAS_QT = True
+except Exception:  # pragma: no cover - PyQt5 not installed
+    _HAS_QT = False
+
+
+@unittest.skipUnless(_HAS_QT, "PyQt5 not available")
+class TranslationEngineDropdownTests(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        cls._app = QApplication.instance() or QApplication([])
+
+    def test_all_engines_are_listed_without_environment_configuration(self) -> None:
+        required_settings = (
+            "model",
+            "language",
+            "timeout",
+            "width",
+            "listWidth",
+            "noteWidth",
+        )
+        without_settings = {
+            key: value
+            for key, value in os.environ.items()
+            if key not in required_settings
+        }
+
+        with patch.dict(os.environ, without_settings, clear=True):
+            tab = TranslationTab()
+
+        expected = [
+            f"{name} ({', '.join(extensions)})"
+            for name, extensions, _module, _handler in TRANSLATION_MODULE_SPECS
+        ]
+        actual = [
+            tab.module_combo.itemText(index)
+            for index in range(tab.module_combo.count())
+        ]
+        self.assertEqual(actual, expected)
+        self.assertTrue(all(callable(module[2]) for module in tab.modules))
+
+
+if __name__ == "__main__":
+    unittest.main()