health-score
Raised context thresholds + critical coaching update
Raised the HEALTHY_CEIL from 50% to 70% and DEGRADING_CEIL from 80% to 90% based on user feedback that early warnings were firing too soon for experienced users. The degrading zone now starts at 70% and critical at 90%, matching actual model attention curves more closely. Updated the critical coaching message: "Context nearly full. Start a new chat, or use Claude Projects for ongoing work." Added a degrading-boundary integration test at 75% context to prevent regressions. 1,452 tests passing. PR #42.