Implement Verification Feedback Loop for Research Agent

[dixsep]

This PR implements a critical enhancement to our agent workflow by adding a verification feedback loop.

Previously, when the verify step failed, the process would loop back to the research agent without providing any context on why the verification failed. This led to inefficient retries and a high probability of the agent repeating the same error.

This change introduces a feedback mechanism where the specific reason for the verification failure is passed back to the research agent

Changes Implemented :
Modified the agent workflow to capture and pass failure reasons from the verify step to the research step.

Updated the re-research prompt template to include the previous failed answer and the specific verification feedback.

This enables the agent to perform a targeted correction rather than a blind retry.

This new feature will significantly improve the agent's:

Accuracy: By directly addressing specific errors.

Efficiency: By reducing wasted cycles and repeated failures.

Reliability: By creating a more robust, self-correcting system.