fix(auth): read .env as utf-8-sig in the dotenv-vs-shell detector

_remove_env_source() decides whether a credential var lives in ~/.hermes/.env
or the shell by scanning the .env with env_path.read_text(errors="replace") —
no encoding. read_text() with no encoding falls back to the system locale
(cp1252/GBK on Windows) and never strips a BOM.

The canonical .env readers in hermes_cli/config.py all use
encoding="utf-8-sig" precisely because 'users may edit .env in Notepad which
adds one' (a BOM), and doctor.py documents that .env is written as UTF-8
everywhere. This sibling reader diverged: on a Notepad-edited .env the BOM
prefixes the first line, so line.strip().startswith(f"{env_var}=") is False
for the first variable — the detector reports a .env-backed key as a phantom
shell export and prints a misleading 'still set in your shell environment'
hint on .

Match the canonical reader (utf-8-sig + errors=replace). Adds a regression
test with a BOM'd .env.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

agent/credential_sources.py

@@ -162,9 +162,17 @@ def _remove_env_source(provider: str, removed) -> RemovalResult:
     try:
         env_path = get_env_path()
         if env_path.exists():
+            # Read the .env as UTF-8 with BOM tolerance, matching the
+            # canonical reader in hermes_cli/config.py. read_text() with no
+            # encoding falls back to the system locale (cp1252/GBK on Windows)
+            # and never strips a BOM, so a Notepad-edited .env (BOM + non-ASCII
+            # values) would make the first line fail the startswith() check —
+            # misreporting a .env-backed var as a shell export.
             env_in_dotenv = any(
                 line.strip().startswith(f"{env_var}=")
-                for line in env_path.read_text(errors="replace", encoding="utf-8").splitlines()
+                for line in env_path.read_text(
+                    encoding="utf-8-sig", errors="replace"
+                ).splitlines()
             )
     except OSError:
         pass

tests/hermes_cli/test_auth_commands.py

@@ -769,3 +769,44 @@ def test_auth_remove_copilot_suppresses_all_variants(tmp_path, monkeypatch):
     assert is_source_suppressed("copilot", "env:GITHUB_TOKEN")
 
 
+def test_auth_remove_env_seeded_dotenv_with_bom_no_shell_hint(tmp_path, monkeypatch, capsys):
+    """A Notepad-edited .env carries a UTF-8 BOM. The dotenv-vs-shell
+    detector must still see the first variable as living in .env (and not
+    warn about a phantom shell export). Regression for the reader that
+    dropped encoding='utf-8-sig' and misread the BOM'd first line.
+    """
+    hermes_home = tmp_path / "hermes"
+    hermes_home.mkdir(parents=True, exist_ok=True)
+    monkeypatch.setenv("HERMES_HOME", str(hermes_home))
+
+    monkeypatch.delenv("DEEPSEEK_API_KEY", raising=False)
+    # BOM prefix (utf-8-sig) + the target var as the FIRST line.
+    (hermes_home / ".env").write_bytes(
+        b"\xef\xbb\xbfDEEPSEEK_API_KEY=sk-ds-only\n"
+    )
+    monkeypatch.setenv("DEEPSEEK_API_KEY", "sk-ds-only")
+
+    _write_auth_store(
+        tmp_path,
+        {
+            "version": 1,
+            "credential_pool": {
+                "deepseek": [{
+                    "id": "env-1",
+                    "label": "DEEPSEEK_API_KEY",
+                    "auth_type": "api_key",
+                    "priority": 0,
+                    "source": "env:DEEPSEEK_API_KEY",
+                    "access_token": "sk-ds-only",
+                }]
+            },
+        },
+    )
+
+    from types import SimpleNamespace
+    from hermes_cli.auth_commands import auth_remove_command
+    auth_remove_command(SimpleNamespace(provider="deepseek", target="1"))
+
+    out = capsys.readouterr().out
+    assert "Cleared DEEPSEEK_API_KEY from .env" in out
+    assert "still set in your shell environment" not in out