# Task
Validate if the agent's output matches the original task and agent's instructions.

# Original Task
<original_task>
%original_task%
</original_task>

# Agent Prompt (Instructions the agent followed)
<agent_prompt>
%agent_prompt%
</agent_prompt>

# Agent Result
<agent_result>
%agent_result%
</agent_result>

# Validation Rules
1. Check that output matches the expected format from agent prompt
2. Verify data is correctly extracted from original task
3. Check format correctness

# Response Format
Return ONLY valid JSON without any backticks
{
  "approved": true or false,
  "feedback": "What needs to be fixed (only if not approved)"
}