fix: extend JSON repair to browser tool and CLI (#862)

- repair malformed browser command envelopes before treating stdout as non-JSON
- repair eval/image payloads in browser_tool.py
- repair CLI parsing for vision and /cron tool responses
- add focused regression tests for browser and CLI repair paths

agent/anthropic_adapter.py

@@ -1396,8 +1396,6 @@ def normalize_anthropic_response(
         "tool_use": "tool_calls",
         "max_tokens": "length",
         "stop_sequence": "stop",
-        "refusal": "content_filter",
-        "model_context_window_exceeded": "length",
     }
     finish_reason = stop_reason_map.get(response.stop_reason, "stop")
 
@@ -1411,42 +1409,3 @@ def normalize_anthropic_response(
         ),
         finish_reason,
     )
-
-
-def normalize_anthropic_response_v2(
-    response,
-    strip_tool_prefix: bool = False,
-) -> "NormalizedResponse":
-    """Normalize Anthropic response to NormalizedResponse.
-
-    Wraps the existing normalize_anthropic_response() and maps its output
-    to the shared transport types. This allows incremental migration
-    without disturbing the legacy call sites.
-    """
-    from agent.transports.types import NormalizedResponse, build_tool_call
-
-    assistant_msg, finish_reason = normalize_anthropic_response(response, strip_tool_prefix)
-
-    tool_calls = None
-    if assistant_msg.tool_calls:
-        tool_calls = [
-            build_tool_call(
-                id=tc.id,
-                name=tc.function.name,
-                arguments=tc.function.arguments,
-            )
-            for tc in assistant_msg.tool_calls
-        ]
-
-    provider_data = {}
-    if getattr(assistant_msg, "reasoning_details", None):
-        provider_data["reasoning_details"] = assistant_msg.reasoning_details
-
-    return NormalizedResponse(
-        content=assistant_msg.content,
-        tool_calls=tool_calls,
-        finish_reason=finish_reason,
-        reasoning=getattr(assistant_msg, "reasoning", None),
-        usage=None,
-        provider_data=provider_data or None,
-    )

agent/transports/__init__.py

@@ -1,57 +0,0 @@
-"""Transport layer types and registry for provider response normalization.
-
-Usage:
-    from agent.transports import get_transport
-    transport = get_transport("anthropic_messages")
-    result = transport.normalize_response(raw_response)
-"""
-
-from agent.transports.types import (  # noqa: F401
-    NormalizedResponse,
-    ToolCall,
-    Usage,
-    build_tool_call,
-    map_finish_reason,
-)
-
-_REGISTRY: dict = {}
-
-
-def register_transport(api_mode: str, transport_cls: type) -> None:
-    """Register a transport class for an api_mode string."""
-    _REGISTRY[api_mode] = transport_cls
-
-
-def get_transport(api_mode: str):
-    """Get a transport instance for the given api_mode.
-
-    Returns None if no transport is registered for this api_mode.
-    This allows gradual migration — call sites can check for None
-    and fall back to the legacy code path.
-    """
-    if not _REGISTRY:
-        _discover_transports()
-    cls = _REGISTRY.get(api_mode)
-    if cls is None:
-        return None
-    return cls()
-
-
-def _discover_transports() -> None:
-    """Import all transport modules to trigger auto-registration."""
-    try:
-        import agent.transports.anthropic  # noqa: F401
-    except ImportError:
-        pass
-    try:
-        import agent.transports.codex  # noqa: F401
-    except ImportError:
-        pass
-    try:
-        import agent.transports.chat_completions  # noqa: F401
-    except ImportError:
-        pass
-    try:
-        import agent.transports.bedrock  # noqa: F401
-    except ImportError:
-        pass

agent/transports/anthropic.py

@@ -1,95 +0,0 @@
-"""Anthropic Messages API transport.
-
-Delegates to the existing adapter functions in agent/anthropic_adapter.py.
-This transport owns format conversion and normalization — NOT client lifecycle.
-"""
-
-from typing import Any, Dict, List, Optional
-
-from agent.transports.base import ProviderTransport
-from agent.transports.types import NormalizedResponse
-
-
-class AnthropicTransport(ProviderTransport):
-    """Transport for api_mode='anthropic_messages'."""
-
-    @property
-    def api_mode(self) -> str:
-        return "anthropic_messages"
-
-    def convert_messages(self, messages: List[Dict[str, Any]], **kwargs) -> Any:
-        from agent.anthropic_adapter import convert_messages_to_anthropic
-
-        base_url = kwargs.get("base_url")
-        return convert_messages_to_anthropic(messages, base_url=base_url)
-
-    def convert_tools(self, tools: List[Dict[str, Any]]) -> Any:
-        from agent.anthropic_adapter import convert_tools_to_anthropic
-
-        return convert_tools_to_anthropic(tools)
-
-    def build_kwargs(
-        self,
-        model: str,
-        messages: List[Dict[str, Any]],
-        tools: Optional[List[Dict[str, Any]]] = None,
-        **params,
-    ) -> Dict[str, Any]:
-        from agent.anthropic_adapter import build_anthropic_kwargs
-
-        return build_anthropic_kwargs(
-            model=model,
-            messages=messages,
-            tools=tools,
-            max_tokens=params.get("max_tokens", 16384),
-            reasoning_config=params.get("reasoning_config"),
-            tool_choice=params.get("tool_choice"),
-            is_oauth=params.get("is_oauth", False),
-            preserve_dots=params.get("preserve_dots", False),
-            context_length=params.get("context_length"),
-            base_url=params.get("base_url"),
-            fast_mode=params.get("fast_mode", False),
-        )
-
-    def normalize_response(self, response: Any, **kwargs) -> NormalizedResponse:
-        from agent.anthropic_adapter import normalize_anthropic_response_v2
-
-        strip_tool_prefix = kwargs.get("strip_tool_prefix", False)
-        return normalize_anthropic_response_v2(response, strip_tool_prefix=strip_tool_prefix)
-
-    def validate_response(self, response: Any) -> bool:
-        if response is None:
-            return False
-        content_blocks = getattr(response, "content", None)
-        if not isinstance(content_blocks, list):
-            return False
-        if not content_blocks:
-            return False
-        return True
-
-    def extract_cache_stats(self, response: Any):
-        usage = getattr(response, "usage", None)
-        if usage is None:
-            return None
-        cached = getattr(usage, "cache_read_input_tokens", 0) or 0
-        written = getattr(usage, "cache_creation_input_tokens", 0) or 0
-        if cached or written:
-            return {"cached_tokens": cached, "creation_tokens": written}
-        return None
-
-    _STOP_REASON_MAP = {
-        "end_turn": "stop",
-        "tool_use": "tool_calls",
-        "max_tokens": "length",
-        "stop_sequence": "stop",
-        "refusal": "content_filter",
-        "model_context_window_exceeded": "length",
-    }
-
-    def map_finish_reason(self, raw_reason: str) -> str:
-        return self._STOP_REASON_MAP.get(raw_reason, "stop")
-
-
-from agent.transports import register_transport  # noqa: E402
-
-register_transport("anthropic_messages", AnthropicTransport)

agent/transports/base.py

@@ -1,61 +0,0 @@
-"""Abstract base for provider transports.
-
-A transport owns the data path for one api_mode:
-  convert_messages → convert_tools → build_kwargs → normalize_response
-
-It does NOT own: client construction, streaming, credential refresh,
-prompt caching, interrupt handling, or retry logic. Those stay on AIAgent.
-"""
-
-from abc import ABC, abstractmethod
-from typing import Any, Dict, List, Optional
-
-from agent.transports.types import NormalizedResponse
-
-
-class ProviderTransport(ABC):
-    """Base class for provider-specific format conversion and normalization."""
-
-    @property
-    @abstractmethod
-    def api_mode(self) -> str:
-        """The api_mode string this transport handles."""
-        ...
-
-    @abstractmethod
-    def convert_messages(self, messages: List[Dict[str, Any]], **kwargs) -> Any:
-        """Convert OpenAI-format messages to provider-native format."""
-        ...
-
-    @abstractmethod
-    def convert_tools(self, tools: List[Dict[str, Any]]) -> Any:
-        """Convert OpenAI-format tool definitions to provider-native format."""
-        ...
-
-    @abstractmethod
-    def build_kwargs(
-        self,
-        model: str,
-        messages: List[Dict[str, Any]],
-        tools: Optional[List[Dict[str, Any]]] = None,
-        **params,
-    ) -> Dict[str, Any]:
-        """Build the complete provider kwargs dict."""
-        ...
-
-    @abstractmethod
-    def normalize_response(self, response: Any, **kwargs) -> NormalizedResponse:
-        """Normalize a raw provider response to the shared NormalizedResponse type."""
-        ...
-
-    def validate_response(self, response: Any) -> bool:
-        """Optional structural validation for raw responses."""
-        return True
-
-    def extract_cache_stats(self, response: Any) -> Optional[Dict[str, int]]:
-        """Optional cache stats extraction."""
-        return None
-
-    def map_finish_reason(self, raw_reason: str) -> str:
-        """Optional stop-reason mapping. Defaults to passthrough."""
-        return raw_reason

agent/transports/types.py

@@ -1,58 +0,0 @@
-"""Shared types for normalized provider responses."""
-
-from __future__ import annotations
-
-import json
-from dataclasses import dataclass, field
-from typing import Any, Dict, List, Optional
-
-
-@dataclass
-class ToolCall:
-    """A normalized tool call from any provider."""
-
-    id: Optional[str]
-    name: str
-    arguments: str
-    provider_data: Optional[Dict[str, Any]] = field(default=None, repr=False)
-
-
-@dataclass
-class Usage:
-    """Token usage from an API response."""
-
-    prompt_tokens: int = 0
-    completion_tokens: int = 0
-    total_tokens: int = 0
-    cached_tokens: int = 0
-
-
-@dataclass
-class NormalizedResponse:
-    """Normalized API response from any provider."""
-
-    content: Optional[str]
-    tool_calls: Optional[List[ToolCall]]
-    finish_reason: str
-    reasoning: Optional[str] = None
-    usage: Optional[Usage] = None
-    provider_data: Optional[Dict[str, Any]] = field(default=None, repr=False)
-
-
-def build_tool_call(
-    id: Optional[str],
-    name: str,
-    arguments: Any,
-    **provider_fields: Any,
-) -> ToolCall:
-    """Build a ToolCall, auto-serialising dict arguments."""
-    args_str = json.dumps(arguments) if isinstance(arguments, dict) else str(arguments)
-    provider_data = dict(provider_fields) if provider_fields else None
-    return ToolCall(id=id, name=name, arguments=args_str, provider_data=provider_data)
-
-
-def map_finish_reason(reason: Optional[str], mapping: Dict[str, str]) -> str:
-    """Translate a provider-specific stop reason to the normalized set."""
-    if reason is None:
-        return "stop"
-    return mapping.get(reason, "stop")

cli.py

@@ -589,6 +589,7 @@ from tools.terminal_tool import set_sudo_password_callback, set_approval_callbac
 from tools.skills_tool import set_secret_capture_callback
 from hermes_cli.callbacks import prompt_for_secret
 from tools.browser_tool import _emergency_cleanup_all_sessions as _cleanup_all_browsers
+from utils import repair_and_load_json
 
 # Guard to prevent cleanup from running multiple times on exit
 _cleanup_done = False
@@ -3569,7 +3570,11 @@ class HermesCLI:
                 result_json = _asyncio.run(
                     vision_analyze_tool(image_url=str(img_path), user_prompt=analysis_prompt)
                 )
-                result = _json.loads(result_json)
+                result = repair_and_load_json(
+                    result_json,
+                    default={},
+                    context="cli_image_analysis",
+                ) if isinstance(result_json, str) else {}
                 if result.get("success"):
                     description = result.get("analysis", "")
                     enriched_parts.append(
@@ -4960,7 +4965,14 @@ class HermesCLI:
         from tools.cronjob_tools import cronjob as cronjob_tool
 
         def _cron_api(**kwargs):
-            return json.loads(cronjob_tool(**kwargs))
+            result = repair_and_load_json(
+                cronjob_tool(**kwargs),
+                default=None,
+                context="cli_cron_command",
+            )
+            if isinstance(result, dict):
+                return result
+            return {"success": False, "error": "Invalid JSON from cronjob tool"}
 
         def _normalize_skills(values):
             normalized = []

tests/agent/test_anthropic_normalize_v2.py

@@ -1,213 +0,0 @@
-"""Regression tests: normalize_anthropic_response_v2 vs v1.
-
-Constructs mock Anthropic responses and asserts that the v2 function
-(returning NormalizedResponse) produces identical field values to the
-original v1 function (returning SimpleNamespace + finish_reason).
-"""
-
-from types import SimpleNamespace
-
-import pytest
-
-from agent.anthropic_adapter import (
-    normalize_anthropic_response,
-    normalize_anthropic_response_v2,
-)
-from agent.transports.types import NormalizedResponse
-
-
-def _text_block(text: str):
-    return SimpleNamespace(type="text", text=text)
-
-
-def _thinking_block(thinking: str, signature: str = "sig_abc"):
-    return SimpleNamespace(type="thinking", thinking=thinking, signature=signature)
-
-
-def _tool_use_block(id: str, name: str, input: dict):
-    return SimpleNamespace(type="tool_use", id=id, name=name, input=input)
-
-
-def _response(content_blocks, stop_reason="end_turn"):
-    return SimpleNamespace(
-        content=content_blocks,
-        stop_reason=stop_reason,
-        usage=SimpleNamespace(input_tokens=10, output_tokens=5),
-    )
-
-
-class TestTextOnly:
-    def setup_method(self):
-        self.resp = _response([_text_block("Hello world")])
-        self.v1_msg, self.v1_finish = normalize_anthropic_response(self.resp)
-        self.v2 = normalize_anthropic_response_v2(self.resp)
-
-    def test_type(self):
-        assert isinstance(self.v2, NormalizedResponse)
-
-    def test_content_matches(self):
-        assert self.v2.content == self.v1_msg.content
-
-    def test_finish_reason_matches(self):
-        assert self.v2.finish_reason == self.v1_finish
-
-    def test_no_tool_calls(self):
-        assert self.v2.tool_calls is None
-        assert self.v1_msg.tool_calls is None
-
-    def test_no_reasoning(self):
-        assert self.v2.reasoning is None
-        assert self.v1_msg.reasoning is None
-
-
-class TestWithToolCalls:
-    def setup_method(self):
-        self.resp = _response(
-            [
-                _text_block("I'll check that"),
-                _tool_use_block("toolu_abc", "terminal", {"command": "ls"}),
-                _tool_use_block("toolu_def", "read_file", {"path": "/tmp"}),
-            ],
-            stop_reason="tool_use",
-        )
-        self.v1_msg, self.v1_finish = normalize_anthropic_response(self.resp)
-        self.v2 = normalize_anthropic_response_v2(self.resp)
-
-    def test_finish_reason(self):
-        assert self.v2.finish_reason == "tool_calls"
-        assert self.v1_finish == "tool_calls"
-
-    def test_tool_call_count(self):
-        assert len(self.v2.tool_calls) == 2
-        assert len(self.v1_msg.tool_calls) == 2
-
-    def test_tool_call_ids_match(self):
-        for i in range(2):
-            assert self.v2.tool_calls[i].id == self.v1_msg.tool_calls[i].id
-
-    def test_tool_call_names_match(self):
-        assert self.v2.tool_calls[0].name == "terminal"
-        assert self.v2.tool_calls[1].name == "read_file"
-        for i in range(2):
-            assert self.v2.tool_calls[i].name == self.v1_msg.tool_calls[i].function.name
-
-    def test_tool_call_arguments_match(self):
-        for i in range(2):
-            assert self.v2.tool_calls[i].arguments == self.v1_msg.tool_calls[i].function.arguments
-
-    def test_content_preserved(self):
-        assert self.v2.content == self.v1_msg.content
-        assert "check that" in self.v2.content
-
-
-class TestWithThinking:
-    def setup_method(self):
-        self.resp = _response([
-            _thinking_block("Let me think about this carefully..."),
-            _text_block("The answer is 42."),
-        ])
-        self.v1_msg, self.v1_finish = normalize_anthropic_response(self.resp)
-        self.v2 = normalize_anthropic_response_v2(self.resp)
-
-    def test_reasoning_matches(self):
-        assert self.v2.reasoning == self.v1_msg.reasoning
-        assert "think about this" in self.v2.reasoning
-
-    def test_reasoning_details_in_provider_data(self):
-        v1_details = self.v1_msg.reasoning_details
-        v2_details = self.v2.provider_data.get("reasoning_details") if self.v2.provider_data else None
-        assert v1_details is not None
-        assert v2_details is not None
-        assert len(v2_details) == len(v1_details)
-
-    def test_content_excludes_thinking(self):
-        assert self.v2.content == "The answer is 42."
-
-
-class TestMixed:
-    def setup_method(self):
-        self.resp = _response(
-            [
-                _thinking_block("Planning my approach..."),
-                _text_block("I'll run the command"),
-                _tool_use_block("toolu_xyz", "terminal", {"command": "pwd"}),
-            ],
-            stop_reason="tool_use",
-        )
-        self.v1_msg, self.v1_finish = normalize_anthropic_response(self.resp)
-        self.v2 = normalize_anthropic_response_v2(self.resp)
-
-    def test_all_fields_present(self):
-        assert self.v2.content is not None
-        assert self.v2.tool_calls is not None
-        assert self.v2.reasoning is not None
-        assert self.v2.finish_reason == "tool_calls"
-
-    def test_content_matches(self):
-        assert self.v2.content == self.v1_msg.content
-
-    def test_reasoning_matches(self):
-        assert self.v2.reasoning == self.v1_msg.reasoning
-
-    def test_tool_call_matches(self):
-        assert self.v2.tool_calls[0].id == self.v1_msg.tool_calls[0].id
-        assert self.v2.tool_calls[0].name == self.v1_msg.tool_calls[0].function.name
-
-
-class TestStopReasons:
-    @pytest.mark.parametrize("stop_reason,expected", [
-        ("end_turn", "stop"),
-        ("tool_use", "tool_calls"),
-        ("max_tokens", "length"),
-        ("stop_sequence", "stop"),
-        ("refusal", "content_filter"),
-        ("model_context_window_exceeded", "length"),
-        ("unknown_future_reason", "stop"),
-    ])
-    def test_stop_reason_mapping(self, stop_reason, expected):
-        resp = _response([_text_block("x")], stop_reason=stop_reason)
-        _v1_msg, v1_finish = normalize_anthropic_response(resp)
-        v2 = normalize_anthropic_response_v2(resp)
-        assert v2.finish_reason == v1_finish == expected
-
-
-class TestStripToolPrefix:
-    def test_prefix_stripped(self):
-        resp = _response(
-            [_tool_use_block("toolu_1", "mcp_terminal", {"cmd": "ls"})],
-            stop_reason="tool_use",
-        )
-        v1_msg, _ = normalize_anthropic_response(resp, strip_tool_prefix=True)
-        v2 = normalize_anthropic_response_v2(resp, strip_tool_prefix=True)
-        assert v1_msg.tool_calls[0].function.name == "terminal"
-        assert v2.tool_calls[0].name == "terminal"
-
-    def test_prefix_kept(self):
-        resp = _response(
-            [_tool_use_block("toolu_1", "mcp_terminal", {"cmd": "ls"})],
-            stop_reason="tool_use",
-        )
-        v1_msg, _ = normalize_anthropic_response(resp, strip_tool_prefix=False)
-        v2 = normalize_anthropic_response_v2(resp, strip_tool_prefix=False)
-        assert v1_msg.tool_calls[0].function.name == "mcp_terminal"
-        assert v2.tool_calls[0].name == "mcp_terminal"
-
-
-class TestEdgeCases:
-    def test_empty_content_blocks(self):
-        resp = _response([])
-        v1_msg, _v1_finish = normalize_anthropic_response(resp)
-        v2 = normalize_anthropic_response_v2(resp)
-        assert v2.content == v1_msg.content
-        assert v2.content is None
-
-    def test_no_reasoning_details_means_none_provider_data(self):
-        resp = _response([_text_block("hi")])
-        v2 = normalize_anthropic_response_v2(resp)
-        assert v2.provider_data is None
-
-    def test_v2_returns_dataclass_not_namespace(self):
-        resp = _response([_text_block("hi")])
-        v2 = normalize_anthropic_response_v2(resp)
-        assert isinstance(v2, NormalizedResponse)
-        assert not isinstance(v2, SimpleNamespace)

tests/agent/transports/test_transport.py

@@ -1,208 +0,0 @@
-"""Tests for the transport ABC, registry, and AnthropicTransport."""
-
-from types import SimpleNamespace
-
-import pytest
-
-from agent.transports import _REGISTRY, get_transport, register_transport
-from agent.transports.base import ProviderTransport
-from agent.transports.types import NormalizedResponse
-
-
-class TestProviderTransportABC:
-    def test_cannot_instantiate_abc(self):
-        with pytest.raises(TypeError):
-            ProviderTransport()
-
-    def test_concrete_must_implement_all_abstract(self):
-        class Incomplete(ProviderTransport):
-            @property
-            def api_mode(self):
-                return "test"
-
-        with pytest.raises(TypeError):
-            Incomplete()
-
-    def test_minimal_concrete(self):
-        class Minimal(ProviderTransport):
-            @property
-            def api_mode(self):
-                return "test_minimal"
-
-            def convert_messages(self, messages, **kw):
-                return messages
-
-            def convert_tools(self, tools):
-                return tools
-
-            def build_kwargs(self, model, messages, tools=None, **params):
-                return {"model": model, "messages": messages}
-
-            def normalize_response(self, response, **kw):
-                return NormalizedResponse(content="ok", tool_calls=None, finish_reason="stop")
-
-        t = Minimal()
-        assert t.api_mode == "test_minimal"
-        assert t.validate_response(None) is True
-        assert t.extract_cache_stats(None) is None
-        assert t.map_finish_reason("end_turn") == "end_turn"
-
-
-class TestTransportRegistry:
-    def test_get_unregistered_returns_none(self):
-        assert get_transport("nonexistent_mode") is None
-
-    def test_anthropic_registered_on_import(self):
-        import agent.transports.anthropic  # noqa: F401
-
-        t = get_transport("anthropic_messages")
-        assert t is not None
-        assert t.api_mode == "anthropic_messages"
-
-    def test_register_and_get(self):
-        class DummyTransport(ProviderTransport):
-            @property
-            def api_mode(self):
-                return "dummy_test"
-
-            def convert_messages(self, messages, **kw):
-                return messages
-
-            def convert_tools(self, tools):
-                return tools
-
-            def build_kwargs(self, model, messages, tools=None, **params):
-                return {}
-
-            def normalize_response(self, response, **kw):
-                return NormalizedResponse(content=None, tool_calls=None, finish_reason="stop")
-
-        register_transport("dummy_test", DummyTransport)
-        t = get_transport("dummy_test")
-        assert t.api_mode == "dummy_test"
-        _REGISTRY.pop("dummy_test", None)
-
-
-class TestAnthropicTransport:
-    @pytest.fixture
-    def transport(self):
-        import agent.transports.anthropic  # noqa: F401
-
-        return get_transport("anthropic_messages")
-
-    def test_api_mode(self, transport):
-        assert transport.api_mode == "anthropic_messages"
-
-    def test_convert_tools_simple(self, transport):
-        tools = [{
-            "type": "function",
-            "function": {
-                "name": "test_tool",
-                "description": "A test",
-                "parameters": {"type": "object", "properties": {}},
-            },
-        }]
-        result = transport.convert_tools(tools)
-        assert len(result) == 1
-        assert result[0]["name"] == "test_tool"
-        assert "input_schema" in result[0]
-
-    def test_validate_response_none(self, transport):
-        assert transport.validate_response(None) is False
-
-    def test_validate_response_empty_content(self, transport):
-        r = SimpleNamespace(content=[])
-        assert transport.validate_response(r) is False
-
-    def test_validate_response_valid(self, transport):
-        r = SimpleNamespace(content=[SimpleNamespace(type="text", text="hello")])
-        assert transport.validate_response(r) is True
-
-    def test_map_finish_reason(self, transport):
-        assert transport.map_finish_reason("end_turn") == "stop"
-        assert transport.map_finish_reason("tool_use") == "tool_calls"
-        assert transport.map_finish_reason("max_tokens") == "length"
-        assert transport.map_finish_reason("stop_sequence") == "stop"
-        assert transport.map_finish_reason("refusal") == "content_filter"
-        assert transport.map_finish_reason("model_context_window_exceeded") == "length"
-        assert transport.map_finish_reason("unknown") == "stop"
-
-    def test_extract_cache_stats_none_usage(self, transport):
-        r = SimpleNamespace(usage=None)
-        assert transport.extract_cache_stats(r) is None
-
-    def test_extract_cache_stats_with_cache(self, transport):
-        usage = SimpleNamespace(cache_read_input_tokens=100, cache_creation_input_tokens=50)
-        r = SimpleNamespace(usage=usage)
-        result = transport.extract_cache_stats(r)
-        assert result == {"cached_tokens": 100, "creation_tokens": 50}
-
-    def test_extract_cache_stats_zero(self, transport):
-        usage = SimpleNamespace(cache_read_input_tokens=0, cache_creation_input_tokens=0)
-        r = SimpleNamespace(usage=usage)
-        assert transport.extract_cache_stats(r) is None
-
-    def test_normalize_response_text(self, transport):
-        r = SimpleNamespace(
-            content=[SimpleNamespace(type="text", text="Hello world")],
-            stop_reason="end_turn",
-            usage=SimpleNamespace(input_tokens=10, output_tokens=5),
-            model="claude-sonnet-4-6",
-        )
-        nr = transport.normalize_response(r)
-        assert isinstance(nr, NormalizedResponse)
-        assert nr.content == "Hello world"
-        assert nr.tool_calls is None or nr.tool_calls == []
-        assert nr.finish_reason == "stop"
-
-    def test_normalize_response_tool_calls(self, transport):
-        r = SimpleNamespace(
-            content=[
-                SimpleNamespace(type="tool_use", id="toolu_123", name="terminal", input={"command": "ls"}),
-            ],
-            stop_reason="tool_use",
-            usage=SimpleNamespace(input_tokens=10, output_tokens=20),
-            model="claude-sonnet-4-6",
-        )
-        nr = transport.normalize_response(r)
-        assert nr.finish_reason == "tool_calls"
-        assert len(nr.tool_calls) == 1
-        tc = nr.tool_calls[0]
-        assert tc.name == "terminal"
-        assert tc.id == "toolu_123"
-        assert '"command"' in tc.arguments
-
-    def test_normalize_response_thinking(self, transport):
-        r = SimpleNamespace(
-            content=[
-                SimpleNamespace(type="thinking", thinking="Let me think..."),
-                SimpleNamespace(type="text", text="The answer is 42"),
-            ],
-            stop_reason="end_turn",
-            usage=SimpleNamespace(input_tokens=10, output_tokens=15),
-            model="claude-sonnet-4-6",
-        )
-        nr = transport.normalize_response(r)
-        assert nr.content == "The answer is 42"
-        assert nr.reasoning == "Let me think..."
-
-    def test_build_kwargs_returns_dict(self, transport):
-        messages = [{"role": "user", "content": "Hello"}]
-        kw = transport.build_kwargs(
-            model="claude-sonnet-4-6",
-            messages=messages,
-            max_tokens=1024,
-        )
-        assert isinstance(kw, dict)
-        assert "model" in kw
-        assert "max_tokens" in kw
-        assert "messages" in kw
-
-    def test_convert_messages_extracts_system(self, transport):
-        messages = [
-            {"role": "system", "content": "You are helpful."},
-            {"role": "user", "content": "Hi"},
-        ]
-        system, msgs = transport.convert_messages(messages)
-        assert system is not None
-        assert len(msgs) >= 1

tests/agent/transports/test_types.py

@@ -1,130 +0,0 @@
-"""Tests for agent/transports/types.py — dataclass construction + helpers."""
-
-import json
-
-from agent.transports.types import (
-    NormalizedResponse,
-    ToolCall,
-    Usage,
-    build_tool_call,
-    map_finish_reason,
-)
-
-
-class TestToolCall:
-    def test_basic_construction(self):
-        tc = ToolCall(id="call_abc", name="terminal", arguments='{"cmd": "ls"}')
-        assert tc.id == "call_abc"
-        assert tc.name == "terminal"
-        assert tc.arguments == '{"cmd": "ls"}'
-        assert tc.provider_data is None
-
-    def test_none_id(self):
-        tc = ToolCall(id=None, name="read_file", arguments="{}")
-        assert tc.id is None
-
-    def test_provider_data(self):
-        tc = ToolCall(
-            id="call_x",
-            name="t",
-            arguments="{}",
-            provider_data={"call_id": "call_x", "response_item_id": "fc_x"},
-        )
-        assert tc.provider_data["call_id"] == "call_x"
-        assert tc.provider_data["response_item_id"] == "fc_x"
-
-
-class TestUsage:
-    def test_defaults(self):
-        u = Usage()
-        assert u.prompt_tokens == 0
-        assert u.completion_tokens == 0
-        assert u.total_tokens == 0
-        assert u.cached_tokens == 0
-
-    def test_explicit(self):
-        u = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150, cached_tokens=80)
-        assert u.total_tokens == 150
-
-
-class TestNormalizedResponse:
-    def test_text_only(self):
-        r = NormalizedResponse(content="hello", tool_calls=None, finish_reason="stop")
-        assert r.content == "hello"
-        assert r.tool_calls is None
-        assert r.finish_reason == "stop"
-        assert r.reasoning is None
-        assert r.usage is None
-        assert r.provider_data is None
-
-    def test_with_tool_calls(self):
-        tcs = [ToolCall(id="call_1", name="terminal", arguments='{"cmd":"pwd"}')]
-        r = NormalizedResponse(content=None, tool_calls=tcs, finish_reason="tool_calls")
-        assert r.finish_reason == "tool_calls"
-        assert len(r.tool_calls) == 1
-        assert r.tool_calls[0].name == "terminal"
-
-    def test_with_reasoning(self):
-        r = NormalizedResponse(
-            content="answer",
-            tool_calls=None,
-            finish_reason="stop",
-            reasoning="I thought about it",
-        )
-        assert r.reasoning == "I thought about it"
-
-    def test_with_provider_data(self):
-        r = NormalizedResponse(
-            content=None,
-            tool_calls=None,
-            finish_reason="stop",
-            provider_data={"reasoning_details": [{"type": "thinking", "thinking": "hmm"}]},
-        )
-        assert r.provider_data["reasoning_details"][0]["type"] == "thinking"
-
-
-class TestBuildToolCall:
-    def test_dict_arguments_serialized(self):
-        tc = build_tool_call(id="call_1", name="terminal", arguments={"cmd": "ls"})
-        assert tc.arguments == json.dumps({"cmd": "ls"})
-        assert tc.provider_data is None
-
-    def test_string_arguments_passthrough(self):
-        tc = build_tool_call(id="call_2", name="read_file", arguments='{"path": "/tmp"}')
-        assert tc.arguments == '{"path": "/tmp"}'
-
-    def test_provider_fields(self):
-        tc = build_tool_call(
-            id="call_3",
-            name="terminal",
-            arguments="{}",
-            call_id="call_3",
-            response_item_id="fc_3",
-        )
-        assert tc.provider_data == {"call_id": "call_3", "response_item_id": "fc_3"}
-
-    def test_none_id(self):
-        tc = build_tool_call(id=None, name="t", arguments="{}")
-        assert tc.id is None
-
-
-class TestMapFinishReason:
-    ANTHROPIC_MAP = {
-        "end_turn": "stop",
-        "tool_use": "tool_calls",
-        "max_tokens": "length",
-        "stop_sequence": "stop",
-        "refusal": "content_filter",
-    }
-
-    def test_known_reason(self):
-        assert map_finish_reason("end_turn", self.ANTHROPIC_MAP) == "stop"
-        assert map_finish_reason("tool_use", self.ANTHROPIC_MAP) == "tool_calls"
-        assert map_finish_reason("max_tokens", self.ANTHROPIC_MAP) == "length"
-        assert map_finish_reason("refusal", self.ANTHROPIC_MAP) == "content_filter"
-
-    def test_unknown_reason_defaults_to_stop(self):
-        assert map_finish_reason("something_new", self.ANTHROPIC_MAP) == "stop"
-
-    def test_none_reason(self):
-        assert map_finish_reason(None, self.ANTHROPIC_MAP) == "stop"

tests/cli/test_cli_json_repair.py

@@ -0,0 +1,62 @@
+import sys
+import types
+from unittest.mock import patch
+
+
+def _stub_auxiliary_client():
+    stub = types.ModuleType("agent.auxiliary_client")
+    stub.call_llm = lambda *args, **kwargs: None
+    stub.resolve_provider_client = lambda *args, **kwargs: (None, None)
+    stub.get_text_auxiliary_client = lambda *args, **kwargs: (None, None)
+    stub.async_call_llm = lambda *args, **kwargs: None
+    stub.extract_content_or_reasoning = lambda *args, **kwargs: ""
+    stub._OR_HEADERS = {}
+    stub._get_task_timeout = lambda *args, **kwargs: 30
+    sys.modules["agent.auxiliary_client"] = stub
+
+
+def _stub_vision_tools(vision_analyze_tool):
+    stub = types.ModuleType("tools.vision_tools")
+    stub.vision_analyze_tool = vision_analyze_tool
+    sys.modules["tools.vision_tools"] = stub
+
+
+def test_preprocess_images_with_vision_repairs_malformed_json(tmp_path):
+    _stub_auxiliary_client()
+    from cli import HermesCLI
+
+    cli_obj = HermesCLI.__new__(HermesCLI)
+    image_path = tmp_path / "test.png"
+    image_path.write_bytes(b"fake-image-bytes")
+
+    async def fake_vision(**kwargs):
+        return "{'success': true, 'analysis': 'Recovered image description',}"
+
+    _stub_vision_tools(fake_vision)
+    result = HermesCLI._preprocess_images_with_vision(
+        cli_obj,
+        "Describe this",
+        [image_path],
+        announce=False,
+    )
+
+    assert "Recovered image description" in result
+    assert "Describe this" in result
+    assert str(image_path) in result
+
+
+def test_handle_cron_command_repairs_malformed_json(capsys):
+    _stub_auxiliary_client()
+    from cli import HermesCLI
+
+    cli_obj = HermesCLI.__new__(HermesCLI)
+    malformed_result = """{'success': true, 'jobs': [{'job_id': 'job-1234567890ab', 'name': 'Nightly Check', 'state': 'scheduled', 'schedule': 'every 1h', 'repeat': 'forever', 'prompt_preview': 'Check server status', 'skills': ['blogwatcher',], 'next_run_at': '2026-04-22T01:00:00Z',},],}"""
+
+    with patch("tools.cronjob_tools.cronjob", return_value=malformed_result):
+        HermesCLI._handle_cron_command(cli_obj, "/cron list")
+
+    out = capsys.readouterr().out
+    assert "Scheduled Jobs:" in out
+    assert "job-1234567890ab" in out
+    assert "Nightly Check" in out
+    assert "blogwatcher" in out

tests/tools/test_browser_json_repair.py

@@ -0,0 +1,108 @@
+import io
+import json
+import sys
+import types
+from unittest.mock import MagicMock, patch
+
+
+def _stub_auxiliary_client():
+    stub = types.ModuleType("agent.auxiliary_client")
+    stub.call_llm = lambda *args, **kwargs: None
+    stub.resolve_provider_client = lambda *args, **kwargs: (None, None)
+    stub.get_text_auxiliary_client = lambda *args, **kwargs: (None, None)
+    stub.async_call_llm = lambda *args, **kwargs: None
+    stub.extract_content_or_reasoning = lambda *args, **kwargs: ""
+    stub._OR_HEADERS = {}
+    stub._get_task_timeout = lambda *args, **kwargs: 30
+    sys.modules["agent.auxiliary_client"] = stub
+
+
+def test_run_browser_command_repairs_malformed_stdout_envelope(tmp_path):
+    _stub_auxiliary_client()
+    from tools.browser_tool import _run_browser_command
+
+    mock_proc = MagicMock()
+    mock_proc.returncode = 0
+    mock_proc.wait.return_value = 0
+    fake_session = {
+        "session_name": "test-session",
+        "session_id": "test-id",
+        "cdp_url": None,
+    }
+    malformed_stdout = "{'success': true, 'data': {'url': 'https://example.com',},}"
+
+    def fake_open(path, mode="r", *args, **kwargs):
+        path = str(path)
+        if path.endswith("_stdout_navigate"):
+            return io.StringIO(malformed_stdout)
+        if path.endswith("_stderr_navigate"):
+            return io.StringIO("")
+        raise FileNotFoundError(path)
+
+    with (
+        patch("tools.browser_tool._find_agent_browser", return_value="/usr/bin/agent-browser"),
+        patch("tools.browser_tool._get_session_info", return_value=fake_session),
+        patch("tools.browser_tool._socket_safe_tmpdir", return_value=str(tmp_path)),
+        patch("tools.browser_tool._merge_browser_path", side_effect=lambda p: p),
+        patch("tools.interrupt.is_interrupted", return_value=False),
+        patch("subprocess.Popen", return_value=mock_proc),
+        patch("os.open", return_value=99),
+        patch("os.close"),
+        patch("os.unlink"),
+        patch("builtins.open", side_effect=fake_open),
+    ):
+        result = _run_browser_command("task-1", "navigate", ["https://example.com"])
+
+    assert result["success"] is True
+    assert result["data"]["url"] == "https://example.com"
+
+
+def test_agent_browser_eval_repairs_malformed_json_result():
+    _stub_auxiliary_client()
+    from tools.browser_tool import _browser_eval
+
+    with patch(
+        "tools.browser_tool._run_browser_command",
+        return_value={"success": True, "data": {"result": "{'items': ['a', 'b',],}"}},
+    ):
+        result = json.loads(_browser_eval("document.body.innerText", task_id="test"))
+
+    assert result["success"] is True
+    assert result["result"] == {"items": ["a", "b"]}
+    assert result["result_type"] == "dict"
+
+
+def test_camofox_eval_repairs_malformed_json_result():
+    _stub_auxiliary_client()
+    from tools.browser_tool import _camofox_eval
+
+    with (
+        patch("tools.browser_camofox._ensure_tab", return_value={"tab_id": "tab-1", "user_id": "user-1"}),
+        patch("tools.browser_camofox._post", return_value={"result": "{'count': 3,}"}),
+    ):
+        result = json.loads(_camofox_eval("2+1", task_id="test"))
+
+    assert result["success"] is True
+    assert result["result"] == {"count": 3}
+    assert result["result_type"] == "dict"
+
+
+def test_browser_get_images_repairs_malformed_json_result():
+    _stub_auxiliary_client()
+    from tools.browser_tool import browser_get_images
+
+    with patch(
+        "tools.browser_tool._run_browser_command",
+        return_value={
+            "success": True,
+            "data": {
+                "result": "[{\"src\": \"https://example.com/cat.png\", \"alt\": \"cat\",}]"
+            },
+        },
+    ):
+        result = json.loads(browser_get_images(task_id="test"))
+
+    assert result["success"] is True
+    assert result["count"] == 1
+    assert result["images"] == [{"src": "https://example.com/cat.png", "alt": "cat"}]
+    assert "warning" not in result

tools/browser_tool.py

@@ -67,6 +67,7 @@ from typing import Dict, Any, Optional, List
 from pathlib import Path
 from agent.auxiliary_client import call_llm
 from hermes_constants import get_hermes_home
+from utils import repair_and_load_json
 
 try:
     from tools.website_policy import check_website_access
@@ -1171,8 +1172,12 @@ def _run_browser_command(
             return {"success": False, "error": f"Browser command '{command}' returned no output"}
 
         if stdout_text:
-            try:
-                parsed = json.loads(stdout_text)
+            parsed = repair_and_load_json(
+                stdout_text,
+                default=None,
+                context=f"browser_{command}_stdout",
+            )
+            if isinstance(parsed, dict):
                 # Warn if snapshot came back empty (common sign of daemon/CDP issues)
                 if command == "snapshot" and parsed.get("success"):
                     snap_data = parsed.get("data", {})
@@ -1181,35 +1186,35 @@ def _run_browser_command(
                                        "Possible stale daemon or CDP connection issue. "
                                        "returncode=%s", returncode)
                 return parsed
-            except json.JSONDecodeError:
-                raw = stdout_text[:2000]
-                logger.warning("browser '%s' returned non-JSON output (rc=%s): %s",
-                               command, returncode, raw[:500])
 
-                if command == "screenshot":
-                    stderr_text = (stderr or "").strip()
-                    combined_text = "\n".join(
-                        part for part in [stdout_text, stderr_text] if part
+            raw = stdout_text[:2000]
+            logger.warning("browser '%s' returned non-JSON output (rc=%s): %s",
+                           command, returncode, raw[:500])
+
+            if command == "screenshot":
+                stderr_text = (stderr or "").strip()
+                combined_text = "\n".join(
+                    part for part in [stdout_text, stderr_text] if part
+                )
+                recovered_path = _extract_screenshot_path_from_text(combined_text)
+
+                if recovered_path and Path(recovered_path).exists():
+                    logger.info(
+                        "browser 'screenshot' recovered file from non-JSON output: %s",
+                        recovered_path,
                     )
-                    recovered_path = _extract_screenshot_path_from_text(combined_text)
+                    return {
+                        "success": True,
+                        "data": {
+                            "path": recovered_path,
+                            "raw": raw,
+                        },
+                    }
 
-                    if recovered_path and Path(recovered_path).exists():
-                        logger.info(
-                            "browser 'screenshot' recovered file from non-JSON output: %s",
-                            recovered_path,
-                        )
-                        return {
-                            "success": True,
-                            "data": {
-                                "path": recovered_path,
-                                "raw": raw,
-                            },
-                        }
-
-                return {
-                    "success": False,
-                    "error": f"Non-JSON output from agent-browser for '{command}': {raw}"
-                }
+            return {
+                "success": False,
+                "error": f"Non-JSON output from agent-browser for '{command}': {raw}"
+            }
         
         # Check for errors
         if returncode != 0:
@@ -1777,10 +1782,11 @@ def _browser_eval(expression: str, task_id: Optional[str] = None) -> str:
     # is valid JSON, parse it so the model gets structured data.
     parsed = raw_result
     if isinstance(raw_result, str):
-        try:
-            parsed = json.loads(raw_result)
-        except (json.JSONDecodeError, ValueError):
-            pass  # keep as string
+        parsed = repair_and_load_json(
+            raw_result,
+            default=raw_result,
+            context="browser_eval_result",
+        )
 
     return json.dumps({
         "success": True,
@@ -1801,10 +1807,11 @@ def _camofox_eval(expression: str, task_id: Optional[str] = None) -> str:
         raw_result = resp.get("result") if isinstance(resp, dict) else resp
         parsed = raw_result
         if isinstance(raw_result, str):
-            try:
-                parsed = json.loads(raw_result)
-            except (json.JSONDecodeError, ValueError):
-                pass
+            parsed = repair_and_load_json(
+                raw_result,
+                default=raw_result,
+                context="camofox_eval_result",
+            )
 
         return json.dumps({
             "success": True,
@@ -1904,26 +1911,29 @@ def browser_get_images(task_id: Optional[str] = None) -> str:
     if result.get("success"):
         data = result.get("data", {})
         raw_result = data.get("result", "[]")
-        
-        try:
-            # Parse the JSON string returned by JavaScript
-            if isinstance(raw_result, str):
-                images = json.loads(raw_result)
-            else:
-                images = raw_result
-            
-            return json.dumps({
-                "success": True,
-                "images": images,
-                "count": len(images)
-            }, ensure_ascii=False)
-        except json.JSONDecodeError:
-            return json.dumps({
-                "success": True,
-                "images": [],
-                "count": 0,
-                "warning": "Could not parse image data"
-            }, ensure_ascii=False)
+
+        warning = None
+        if isinstance(raw_result, str):
+            images = repair_and_load_json(
+                raw_result,
+                default=None,
+                context="browser_get_images_result",
+            )
+        else:
+            images = raw_result
+
+        if not isinstance(images, list):
+            images = []
+            warning = "Could not parse image data"
+
+        payload = {
+            "success": True,
+            "images": images,
+            "count": len(images),
+        }
+        if warning:
+            payload["warning"] = warning
+        return json.dumps(payload, ensure_ascii=False)
     else:
         return json.dumps({
             "success": False,