karen P3: sentiment report + multi-business benchmarks

[deepmasq]

Summary

• Implement proper problem reporting functionality in Karen bot JIRA tool #31 Sentiment summary in reports: handle_report now queries done tasks by resolution code (SUCCESS/FAIL/INCONCLUSIVE/ESCALATED) and pre-fills section04-sentiment in the daily/weekly report. Model fills in sentiment_notes if patterns stand out.
• remove superuser from fi_github, fix no output in gh tool, add security check before mongo cat #34 Multi-business benchmarks: Two new scenario YAMLs — B2B fitness SaaS (FitRunner) and fashion e-commerce (Nordic Thread). Tests KB grounding, no-fabrication, and multi-section policy retrieval across business types not yet covered.

Test plan

• Run very_limited__support_fitness_saas scenario on staging
• Run very_limited__support_ecommerce_fashion scenario on staging
• Trigger karen_report(report_type='daily') on a workspace with done tasks and verify section04-sentiment is populated

🤖 Generated with Claude Code