fix(workspace): write and read the vault as UTF-8, not the host locale

anton/cli.py

@@ -345,9 +345,9 @@ def _ensure_terms_consent(console: Console, settings) -> None:
     env_path.parent.mkdir(parents=True, exist_ok=True)
 
     # Append if file exists, otherwise create
-    existing = env_path.read_text() if env_path.is_file() else ""
+    existing = env_path.read_text(encoding="utf-8") if env_path.is_file() else ""
     if "ANTON_TERMS_CONSENT" not in existing:
-        with env_path.open("a") as f:
+        with env_path.open("a", encoding="utf-8") as f:
             if existing and not existing.endswith("\n"):
                 f.write("\n")
             f.write("ANTON_TERMS_CONSENT=true\n")

anton/workspace.py

@@ -120,13 +120,14 @@ def _initialize_once(self, *, create_anton_md: bool = True) -> list[str]:
         # Create anton.md if it doesn't exist
         if create_anton_md and not self._anton_md.is_file():
             self._anton_md.write_text(
-                ANTON_MD_TEMPLATE.format(date=datetime.now().strftime("%Y-%m-%d"))
+                ANTON_MD_TEMPLATE.format(date=datetime.now().strftime("%Y-%m-%d")),
+                encoding="utf-8",
             )
             actions.append(f"Created {self._anton_md}")
 
         # Create .env if it doesn't exist
         if not self._env_file.is_file():
-            self._env_file.write_text("# Anton environment variables\n")
+            self._env_file.write_text("# Anton environment variables\n", encoding="utf-8")
             actions.append(f"Created {self._env_file}")
 
         # Visible artifacts directory at the workspace root. Replaces
@@ -150,7 +151,7 @@ def read_anton_md(self) -> str | None:
         """Read anton.md content. Returns None if it doesn't exist."""
         if not self._anton_md.is_file():
             return None
-        return self._anton_md.read_text()
+        return self._anton_md.read_text(encoding="utf-8")
 
     def anton_md_modified_since_last_read(self) -> bool:
         """Check if anton.md has been modified since last read_anton_md_tracked()."""
@@ -187,7 +188,7 @@ def load_env(self) -> dict[str, str]:
         result: dict[str, str] = {}
         if not self._env_file.is_file():
             return result
-        for line in self._env_file.read_text().splitlines():
+        for line in self._env_file.read_text(encoding="utf-8").splitlines():
             line = line.strip()
             if not line or line.startswith("#"):
                 continue
@@ -217,7 +218,7 @@ def set_secret(self, key: str, value: str) -> None:
         lines: list[str] = []
         replaced = False
         if self._env_file.is_file():
-            for line in self._env_file.read_text().splitlines():
+            for line in self._env_file.read_text(encoding="utf-8").splitlines():
                 stripped = line.strip()
                 if stripped and not stripped.startswith("#") and "=" in stripped:
                     existing_key = stripped.partition("=")[0].strip()
@@ -230,7 +231,7 @@ def set_secret(self, key: str, value: str) -> None:
         if not replaced:
             lines.append(f"{key}={value}")
 
-        self._env_file.write_text("\n".join(lines) + "\n")
+        self._env_file.write_text("\n".join(lines) + "\n", encoding="utf-8")
 
         # Also set in current process environment
         os.environ[key] = value
@@ -245,7 +246,7 @@ def remove_secret(self, key: str) -> bool:
 
         lines: list[str] = []
         found = False
-        for line in self._env_file.read_text().splitlines():
+        for line in self._env_file.read_text(encoding="utf-8").splitlines():
             stripped = line.strip()
             if stripped and not stripped.startswith("#") and "=" in stripped:
                 existing_key = stripped.partition("=")[0].strip()
@@ -255,7 +256,7 @@ def remove_secret(self, key: str) -> bool:
             lines.append(line)
 
         if found:
-            self._env_file.write_text("\n".join(lines) + "\n")
+            self._env_file.write_text("\n".join(lines) + "\n", encoding="utf-8")
             os.environ.pop(key, None)
 
         return found

tests/test_workspace.py

@@ -247,3 +247,77 @@ def test_remove_secret_pops_environ(self, ws, tmp_path):
         assert os.environ.get("ANTON_TEST_REMOVE_XYZ") == "val"
         ws.remove_secret("ANTON_TEST_REMOVE_XYZ")
         assert os.environ.get("ANTON_TEST_REMOVE_XYZ") is None
+
+
+class TestSecretVaultEncoding:
+    # AntonSettings reads .anton/.env with env_file_encoding="utf-8"
+    # (anton/config/settings.py). The vault must therefore be written as
+    # UTF-8, not in the host locale: on a Windows code page (cp1252) a bare
+    # write_text() stores a non-ASCII secret as bytes the settings loader
+    # cannot decode, so every later AntonSettings() raises UnicodeDecodeError.
+
+    def test_non_ascii_secret_is_stored_as_utf8(self, ws):
+        ws.initialize()
+        ws.set_secret("ANTON_MINDS_MIND_NAME", "café_mind")
+
+        raw = ws.env_path.read_bytes()
+        assert "café_mind" in raw.decode("utf-8")
+
+    def test_secret_outside_the_host_code_page_is_storable(self, ws):
+        # 密钥 has no cp1252 representation, so a bare write_text() raises
+        # UnicodeEncodeError and the secret is never stored at all.
+        ws.initialize()
+        ws.set_secret("ANTON_MINDS_DATASOURCE", "密钥")
+        assert ws.get_secret("ANTON_MINDS_DATASOURCE") == "密钥"
+
+    def test_settings_can_load_a_vault_holding_a_non_ascii_secret(self, ws):
+        from anton.config.settings import AntonSettings
+
+        ws.initialize()
+        ws.set_secret("ANTON_MINDS_MIND_NAME", "café_mind")
+        assert AntonSettings(_env_file=str(ws.env_path)).minds_mind_name == "café_mind"
+
+    def test_anton_md_is_read_as_utf8(self, ws):
+        ws.initialize()
+        ws.anton_md_path.write_text("café", encoding="utf-8")
+        assert ws.read_anton_md() == "café"
+
+    def test_workspace_file_io_never_relies_on_the_host_locale(self, tmp_path):
+        # PEP 597: under `-X warn_default_encoding` every locale-default text
+        # open() emits an EncodingWarning. Asserting that none names
+        # workspace.py keeps this regression visible on UTF-8 CI too, where
+        # the behavioural tests above pass whether or not the bug is present.
+        import subprocess
+        import sys
+
+        import anton.workspace as workspace_mod
+
+        repo_root = Path(workspace_mod.__file__).resolve().parent.parent
+        probe = tmp_path / "probe.py"
+        probe.write_text(
+            "import tempfile, warnings\n"
+            "from pathlib import Path\n"
+            "from anton.workspace import Workspace\n"
+            "with warnings.catch_warnings(record=True) as caught:\n"
+            "    warnings.simplefilter('always')\n"
+            "    ws = Workspace(Path(tempfile.mkdtemp()))\n"
+            "    ws.initialize()\n"
+            "    ws.set_secret('K', 'v')\n"
+            "    ws.load_env()\n"
+            "    ws.read_anton_md()\n"
+            "    ws.remove_secret('K')\n"
+            "for w in caught:\n"
+            "    if w.category is EncodingWarning:\n"
+            "        print(f'{w.filename}:{w.lineno}')\n",
+            encoding="utf-8",
+        )
+        proc = subprocess.run(
+            [sys.executable, "-X", "warn_default_encoding", str(probe)],
+            capture_output=True,
+            text=True,
+            env={**os.environ, "PYTHONPATH": str(repo_root)},
+        )
+
+        assert proc.returncode == 0, proc.stderr
+        offenders = [ln for ln in proc.stdout.splitlines() if "workspace.py" in ln]
+        assert offenders == [], f"locale-default text I/O in workspace.py: {offenders}"