fix: update AI governance bullet to be more accurate

fix: swap engagement display order to show client first

feat: add ATS score breakdown with visual indicators

feat: populate ats_score_estimate in resume generation

fix: clamp bullet relevance_score to valid [0, 1] range

fix: filter vague skills from resume (Segment, International, etc.)

fix: improve skills categorization prompt for semantic coherence

fix: use tailored_summary instead of summary in async DOCX generation

feat: implement LLM-powered 3-category skills formatting

fix: expand skills formatter categories for payments/fintech/business

fix: reduce page2_line_budget to 53 for tighter fit

fix: enable pagination-aware condensation by default

fix: correct overflow calculation and add debug logging

chore: reduce page2_line_budget to 20 for testing

chore: reduce page2_line_budget to 30