agentscope-ai/agentscope · error · AgentOrientedException

Input validation failed for tool '{tool_call.name}': {e.mess

Error message

Input validation failed for tool '{tool_call.name}': {e.message}

What it means

Raised as an AgentOrientedException when the arguments parsed for a tool call fail JSON-schema validation against the tool's declared input_schema. The agent rejects malformed LLM-generated tool arguments before invoking the tool function.

Source

Thrown at src/agentscope/agent/_agent.py:2306

        try:
            # Check if the tool is available
            tool = await self.toolkit.check_tool_available(
                tool_call.name,
                self.state.tool_context.activated_groups,
            )

            # Try to parse the input with the tool schema
            parsed_input = _json_loads_with_repair(
                tool_call.input,
                tool.input_schema,
            )

            # Validate the parsed input with the tool schema
            # TODO: Maybe some logic to mix the validation error in runtime
            try:
                jsonschema.validate(parsed_input, tool.input_schema)
            except jsonschema.ValidationError as e:
                raise AgentOrientedException(
                    f"Input validation failed for tool '{tool_call.name}': "
                    f"{e.message}",
                ) from e

        # The exceptions that
        #  - cannot found tool
        #  - tool not available
        #  - input parsing failure
        except AgentOrientedException as e:
            async for evt in self._handle_error_tool_call(
                tool_call,
                e.message,
                state=ToolResultState.ERROR,
            ):
                yield evt

            return

View on GitHub (pinned to e90f1c7592)

Solutions

1. Re-run/let the model retry: the failed validation is reported back to the model so it can correct arguments on the next turn
2. Loosen the tool's input_schema: make optional fields non-required, relax types, or accept extra properties
3. Improve the tool's parameter descriptions and docstring so the model knows expected shapes
4. Use a stronger model if argument hallucination is frequent

Example fix

# before
input_schema = {
    "type": "object",
    "properties": {"path": {"type": "string"}},
    "required": ["path", "mode"],
}

# after
input_schema = {
    "type": "object",
    "properties": {
        "path": {"type": "string", "description": "File path to read"},
        "mode": {"type": "string", "enum": ["r", "rb"], "default": "r"},
    },
    "required": ["path"],
}

import jsonschema

def tool_args_valid(args: dict, tool) -> bool:
    try:
        jsonschema.validate(args, tool.input_schema)
        return True
    except jsonschema.ValidationError:
        return False

from agentscope.exception import AgentOrientedException
try:
    await agent.run(msg)
except AgentOrientedException as e:
    if "Input validation failed" in str(e):
        # feed the error back to the model to self-correct
        msg = UserMsg(f"Tool call rejected: {e}")

When it happens

Trigger: The model emits a tool call whose arguments violate the tool's schema: missing required fields, wrong types (e.g. string where int expected), unknown enum values, or unparseable arguments coerced into an invalid dict.

Common situations: Weak models hallucinating argument names; a tool schema declaring required params the model doesn't reliably supply; overly strict schemas (e.g. strict additionalProperties) after a schema change; prompts not describing parameters clearly.

Related errors

• Invalid structured output from model {model_name}: {e}
• One or more tool calls raised an exception
• The injection template must contain the '{runtime_state}' pl
• Expected a 5-field cron expression, got {record.data.cron_ex
• Invalid values for MCP {card.name!r}: {e.message}

AI-assisted analysis of agentscope-ai/agentscope@e90f1c7592 (2026-08-28). Data as JSON: /api/errors/5284f0388485ee0f. Report an issue: GitHub.