Sprint 19

• compass became the project's operational CLI: db/services/client/ sql pillars (eleven shell scripts deleted), environment status in bearings with DB-drift signalling, sprint audit and sprint charts, and the full PR lifecycle — create, checks, sync, and a guarded merge that closes the task and stamps the journal itself.
• The whole C++ codebase was clang-formatted (16 group passes, CMake targets, nightly workflow) and regrouped under per-group parents; the rfl complexity decomposition finished 16/16, clearing the MSVC C1202 nested-struct limit at the root of the Windows CI break.
• Codegen modernisation: entity models migrated from JSON to org-mode (28/32), C++ and SQL template refactors landed, legacy binary-protocol generator deleted.
• API/core split investigated across all 17 components; the one systemic offender fixed and ores.eventing split into api/core.
• Commission: currency reached 6/8 — codegen, SQL, CLI, shell verified post-NATS; manual chapters and NATS reference in flight.
• The agile system hardened itself: per-concept process pages, five System 2 health reviews, literate sprint charts, and a sprint that closed with zero state-sync violations on its own audit.

Dual-axis bar chart. PRs (left axis) and commits (right axis) per day. A high commits-to-PR ratio may indicate scope creep.

Lines added (green) and deleted (red) per day. Building work produces mostly additions; refactoring produces a mix. Days with no churn may indicate blockers.

Hours from PR open to merge, one bar per PR. Long bars indicate review bottlenecks. Generated only when PR data is available.

Line chart tracking stories marked DONE during the sprint. Steady upward slope is healthy; plateauing signals a stall.

• Throughput and WIP discipline: 186 PRs in nine days, zero open PRs at close, no stale branches — the compass PR lifecycle (guarded merge, automatic task close-out) visibly tightened the loop.
• The sprint audit caught state drift repeatedly and cheaply; the final audit ran clean. Paperwork stopped being archaeology.
• Health reviews worked as designed: HR2 and HR5 both said RED honestly, and HR5's close recommendation is why this sprint is ending deliberately instead of drifting into week three.

• The mission lost to the tooling gravity well — again. The CI goal did get real groundwork (the rfl story cleared the MSVC C1202 root cause) but without a story naming the goal, the health reviews could not see it — and the builds were never re-verified. Goals with no story on day one never got one.
• 30 stories left the sprint unfinished at close. Most were never decomposed — they were planning inventory, not sprint work, and their presence diluted the Stories table's signal all sprint.
• Sprint closure itself was manual and error-prone: the carry mechanics (move vs copy, ID preservation) had to be designed mid-close. Now captured for compass automation.

• Sprint closure now has explicit mechanics: unfinished stories move to the product backlog inbox preserving their immutable IDs (flat story.org for undecomposed stories, whole folders when task history exists), and are marked ABANDONED in the sprint table. Closing and opening sprints are separate PRs.
• Sprint 20 opens with a deliberately narrow mission — the commissioning epic plus codegen tidy-up — and the tooling excuse is spent.