Phase 1: PESTEL Analysis
- Migration 203: pestel_analysis + pestel_factor tables, v_pestel_summary, v_pestel_high_impact views
- Service: create_analysis, add_factor, compute_scores, suggest (auto-populate from platform data), render
- CLI: create, add-factor, list, get, compute, suggest, export
- 20 tests (72 total across all 4 services)

Phase 2: Regional Readiness (Full Analytical Service)
- Migration 204: readiness_dimension, readiness_location_weight, regional_assessment,
  regional_dimension_score, regional_benchmark tables (5 tables, 3 views)
- Seeded 6 dimensions: infrastructure, institutions, market_access, natural_capital,
  policy_environment, human_capital
- Scoring engine: weighted composite scoring, rating bands (A+/A/B/C/D), confidence levels
- Evidence gathering: 7 _query_* functions from existing platform tables/views
- Per-dimension scorers, benchmarking, multi-location comparison
- CLI: create, compute, list, get, compare, create-benchmark, benchmark, dimensions, export
- 20 tests

Phase 3: Publics & Market Segmentation
- Migration 205: stakeholder_public + market_segment tables, v_publics_matrix, v_market_segment_summary
- Stakeholder classification (9 types) with influence/interest scoring and quadrant mapping
- Market segments by buyer type (5 types) with demand patterns
- CLI: add-public, list, update-stance, create-segment, list-segments, map-demand, matrix, suggest
- 18 tests

Phase 4: Environmental Scanning Workflow
- Migration 206: env_scan, env_scan_step, env_scan_source tables (5-step governed lifecycle)
- Steps: identify, gather, analyze, communicate, decide
- Auto-populate from threatcasting, stakeholder feedback, PESTEL, CRISP, market data, advisories
- CLI: create, update-step, get, list, complete, auto-populate, export
- 16 tests

Report Generator: 4 new types (pestel_assessment, regional_readiness, publics_market_landscape, env_scan_report)
AGENTS.md: 40+ new CLI commands documented

_purge() only deleted lifecycle_transition rows for specific test UUIDs, but
predict_remaining() and cycle_time_distribution() read ALL rows for the entity
type. Residual rows from other tests/seed data skewed the median, causing
assertion failures. Now _purge() deletes all lifecycle_transition rows for the
entity type before inserting test data.

Also increased ci-check.sh error output from head -30 to head -60 for better
CI failure visibility.