fix(skills): update MockLLMProvider schema to resolve validation warnings in Online Research Agent

[aarav-shukla07]

Description

This PR resolves the Output validation failed warnings that appear when running the online_research_agent in mock mode.

Previously, the MockLLMProvider returned a generic JSON response that lacked specific keys required by downstream nodes (e.g., fetch-content requires source_urls, write-report requires citations). While the agent would complete execution, the logs were cluttered with validation errors and "Cleaning failed" messages.

This change updates the MockLLMProvider to return a comprehensive "super-schema" that satisfies the input requirements of all nodes in the agent graph.

Type of Change

• Bug fix (non-breaking change that fixes an issue)

Related Issues

Fixes #1223

Changes Made

• Updated MockLLMProvider.complete in agent.py to return a fully populated JSON object.
• Added missing keys required by the graph schema:
  • research_focus
  • source_urls & fetched_sources
  • key_aspects & source_analysis
  • ranked_sources
  • key_findings & themes
  • report_content, source_citations, & references
  • final_report
• Ensured the LLMResponse object wraps the JSON string correctly.

Testing

• Manual testing performed

Manual Verification: Ran the agent in mock mode:

PYTHONPATH=core:exports python3 -m online_research_agent run --topic "AI" --mock

Result

• Execution completed with "success": true.

• Logs are clean of Output validation failed warnings.

• "Haiku formatting failed" warnings may still appear (expected without API key), but schema validation now passes.

Screenshots (if applicable)