whetstone_DSL/sprint89_plan.md

Sprint 89 Plan: Meta-Evaluation and Program Benchmarking

Context

A mature program needs periodic meta-evaluation against external baselines and internal objectives. Sprint 89 adds program-level benchmarking.

---

Goals

1. Define program KPIs for quality, cost, speed, and trust
2. Benchmark program performance over time and across teams
3. Detect systemic regressions not visible at pair level
4. Feed meta-evaluation results into roadmap planning

---

Steps

Step 1119: Program KPI schema and target model (12 tests)

Step 1120: Historical KPI aggregator (10 tests)

Step 1121: Cross-team benchmark comparator (10 tests)

Step 1122: Systemic regression detector (10 tests)

Step 1123: KPI anomaly triage model (10 tests)

Step 1124: Meta-evaluation recommendation generator (8 tests)

Step 1125: whetstone_get_program_kpis MCP tool (8 tests)

Step 1126: whetstone_run_meta_evaluation MCP tool (8 tests)

Step 1127: Program benchmarking report bundle (8 tests)

Step 1128: Sprint 89 integration summary + regression (8 tests)

---

Governance Rule

• Roadmap adjustments must reference KPI evidence from latest meta-evaluation.