feat: Add a self-checking mechanism for tool execution results; If the step execution fails or the self-check is not passed, an attempt will be made to re-execute. After multiple failed executions of the same step, switch to the alternative tool for execution. Update the descriptions of some tools/scenarios

agents/matmaster_agent/flow_agents/scene_agent/model.py

@@ -62,6 +62,8 @@ class SceneEnum(DescriptiveEnum):
     SURFACE_ENERGY = ('surface_energy', '')
     STACKING_FAULT_ENERGY = ('stacking_fault_energy', '')
     EOS = ('eos', '')
+    Electron_Localization_Function = ('electron_localization_function', '')
+    Nudged_Elastic_Band = ('nudged_elastic_band', '')
     PHONON = ('phonon', '')
     SCF = ('scf', 'Self-Consistent Field')
     BAND = ('band', '')

agents/matmaster_agent/flow_agents/step_validation_agent/__init__.py

@@ -0,0 +1,5 @@
+from agents.matmaster_agent.flow_agents.step_validation_agent.agent import (
+    StepValidationAgent,
+)
+
+__all__ = ['StepValidationAgent']

agents/matmaster_agent/flow_agents/step_validation_agent/agent.py

@@ -0,0 +1,30 @@
+import logging
+
+from agents.matmaster_agent.constant import MATMASTER_AGENT_NAME
+from agents.matmaster_agent.core_agents.base_agents.schema_agent import (
+    DisallowTransferAndContentLimitSchemaAgent,
+)
+from agents.matmaster_agent.flow_agents.step_validation_agent.prompt import (
+    STEP_VALIDATION_INSTRUCTION,
+)
+from agents.matmaster_agent.flow_agents.step_validation_agent.schema import (
+    StepValidationSchema,
+)
+from agents.matmaster_agent.llm_config import MatMasterLlmConfig
+from agents.matmaster_agent.logger import PrefixFilter
+
+logger = logging.getLogger(__name__)
+logger.addFilter(PrefixFilter(MATMASTER_AGENT_NAME))
+logger.setLevel(logging.INFO)
+
+
+class StepValidationAgent(DisallowTransferAndContentLimitSchemaAgent):
+    def __init__(self, **kwargs):
+        super().__init__(
+            name='step_validation_agent',
+            model=MatMasterLlmConfig.tool_schema_model,
+            description='校验步骤执行结果是否合理',
+            instruction=STEP_VALIDATION_INSTRUCTION,
+            output_schema=StepValidationSchema,
+            **kwargs,
+        )

agents/matmaster_agent/flow_agents/step_validation_agent/prompt.py

@@ -0,0 +1,41 @@
+STEP_VALIDATION_INSTRUCTION = """
+You are a validation agent responsible for checking if the execution result of a step matches the user's requirements and basic chemical/materials science knowledge.
+
+Your task is to analyze:
+1. The user's original request
+2. The current step's description and purpose
+3. The execution result/output
+4. Basic chemical and materials science principles
+
+Based on this analysis, determine if the result is reasonable and matches expectations.
+
+# Validation Criteria:
+1. **Relevance**: Does the result address the step's intended purpose?
+2. **Accuracy**: Is the result consistent with basic chemical/materials science knowledge?
+3. **Completeness**: Does the result provide the expected information/output?
+4. **Reasonableness**: Are the values, predictions, or conclusions logical?
+
+# Output Format:
+You must respond with a JSON object containing:
+{
+    "is_valid": boolean,  // true if result matches requirements and knowledge, false otherwise
+    "reason": "string",   // brief explanation of validation result
+    "confidence": "high|medium|low"  // confidence level in the validation
+}
+
+# Important Rules:
+- If the result contains obvious errors (wrong chemical formulas, impossible physical properties, etc.), mark as invalid
+- If the result is incomplete or doesn't address the step's purpose, mark as invalid
+- If the result contradicts basic chemical principles, mark as invalid
+- Only mark as valid if the result reasonably matches expectations
+- Be conservative: when in doubt, mark as invalid to ensure quality
+
+# Examples of Invalid Results:
+- Predicting a metal with negative melting point
+- Chemical formula with wrong valence (e.g., NaCl3)
+- Material property values that violate physical laws
+- Results that don't relate to the step's described purpose
+- Incomplete or missing expected outputs
+
+Output only the JSON object, no additional text.
+"""

agents/matmaster_agent/flow_agents/step_validation_agent/schema.py

@@ -0,0 +1,7 @@
+from pydantic import BaseModel
+
+
+class StepValidationSchema(BaseModel):
+    is_valid: bool
+    reason: str
+    confidence: str  # "high", "medium", "low"

agents/matmaster_agent/flow_agents/utils.py

@@ -132,3 +132,17 @@ def should_bypass_confirmation(ctx: InvocationContext) -> bool:
             return True
 
     return False
+
+
+def find_alternative_tool(current_tool_name: str) -> Optional[str]:
+    """Find an alternative tool for the current tool.
+
+    Priority order:
+    1. Pre-defined alternatives in ALL_TOOLS
+    """
+    if current_tool_name not in ALL_TOOLS:
+        return None
+
+    current_tool_info = ALL_TOOLS[current_tool_name]
+
+    return current_tool_info.get('alternative', [])

agents/matmaster_agent/locales.py

@@ -18,6 +18,7 @@
         'RePlan': 'Re-plan with Different Tool(s)',
         'MoreQuestions': 'You may also be interested in these questions:',
         'Step': 'Step',
+        'ReExecuteSteps': 'Re-execute Step',
         'PlanSummary': 'Plan Summary',
         'NoFoundStructure': 'No eligible structures found.',
         'WalletNoFee': 'Insufficient wallet balance',
@@ -40,6 +41,7 @@
         'RePlan': '更换工具重新规划',
         'MoreQuestions': '你或许还对这些问题感兴趣：',
         'Step': '步骤',
+        'ReExecuteSteps': '重试步骤',
         'PlanSummary': '计划汇总概要',
         'NoFoundStructure': '未找到符合条件的结构',
         'WalletNoFee': '钱包余额不足',