Browser-Use MCP Server

"""Skill executor for hint injection and result validation.""" import logging from .models import Skill from .prompts import LEARNING_MODE_SUFFIX, get_execution_hints logger = logging.getLogger(__name__) class SkillExecutor: """Executes skills by injecting hints into agent prompts.""" def inject_hints(self, task: str, skill: Skill, params: dict | None = None) -> str: """Augment task prompt with skill hints. Args: task: Original task description skill: Skill with hints to inject params: Optional parameters to substitute in hints Returns: Augmented task prompt with hints """ params = params or {} # Get formatted hints from skill hints_text = skill.hints.to_prompt(params) if not hints_text.strip(): # No hints to inject, return original task return task # Build execution prompt with hints execution_prompt = get_execution_hints(skill.name, hints_text) return f"{execution_prompt}{task}" def inject_learning_mode(self, task: str) -> str: """Augment task prompt for learning/API discovery mode. Args: task: Original task description Returns: Task with API discovery instructions appended """ return f"{task}\n{LEARNING_MODE_SUFFIX}" def validate_result( self, result: str, skill: Skill, ) -> bool: """Validate execution result against skill expectations. For now, just check that result is non-empty. Future: Compare against expected response schema. Args: result: Agent result string skill: Skill used for execution Returns: True if result appears valid """ if not result or not result.strip(): return False # Basic validation: result exists # TODO: Add schema validation when we have response schemas return True