The agent that grows with you
Python aiai-agentai-agentsanthropicchatgptclaudeclaude-codecodexhermeshermes-agentllmnous-researchopenai
"""`hermes vault sources --disable` must not crash on a hand-edited non-dict
``vault:`` section in config.yaml (the same YAML-shape hazard class
``_voice_cfg_dict`` guards for voice.*, #19835): the malformed section is
coerced to a dict before the opt-out is written."""
from types import SimpleNamespace
import hermes_yaml as yaml
from agent.vault_store import VaultStore
from hermes_cli.vault import _cmd_sources, vault_command
def _corrupt_vault(home):
store = VaultStore(home / "vault")
store.add_item(
"login",
"site",
{"identifier_type": "email", "identifier": "u", "password": "p"},
origin="https://x.example",
)
store._vault_path.write_bytes(store._fernet().encrypt(b"{not json"))
def test_sources_disable_tolerates_scalar_vault_section(tmp_path, monkeypatch):
home = tmp_path / ".hermes"
home.mkdir()
monkeypatch.setenv("HERMES_HOME", str(home))
(home / "config.yaml").write_text("vault: true\n")
_cmd_sources(SimpleNamespace(enable=None, disable="bitwarden"))
saved = yaml.safe_load((home / "config.yaml").read_text())
assert saved["vault"]["bitwarden"]["enabled"] is False
def test_vault_command_surfaces_corrupt_store_as_clean_error(tmp_path, monkeypatch, capsys):
"""A decryptable-but-malformed vault file surfaces through the VaultError
contract as a clean error line on every subcommand, not a traceback."""
home = tmp_path / ".hermes"
home.mkdir()
monkeypatch.setenv("HERMES_HOME", str(home))
_corrupt_vault(home)
vault_command(SimpleNamespace()) # bare `hermes vault` -> _cmd_list
out = capsys.readouterr().out
assert "Error" in out
assert "corrupted" in out