Sprint 20

Stories where work started or partial tasks were completed.

Stories pulled into sprint planning but not touched. No tasks exist; ready to pull into sprint 21.

Every configured build went green simultaneously — all Linux compiler/generator combinations plus the shell and Qt targets — a clean baseline for the provisioning and refdata work in flight.

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 discipline: 161 PRs in six days, closed on day six within the seven-day box — an improvement on sprint 19's nine-day overrun. Zero open PRs at close, no stale branches.
• The compass ecosystem delivered at scale: 31 workflow skills across every pillar, the compass heading next-item recommender, timeline snapshot generation — all three stories complete and shipped.
• The codegen literate migration finished. This commitment carried from sprint 18 through 20 — it is done. All 32+ mustache templates are now literate org sources; MASD knowledge graph fully cross-linked; zero-diff drift check in place.
• Provisioning shipped end-to-end in one sprint: design, CLI implementation, org-mode script library, and a user manual chapter. No half-finished work.
• Sprint audit ran clean at close; health reviews were honest and the HR5 close recommendation was taken without hesitation.

• The commissioning mission repeated the sprint 19 pattern — tooling gravity won. 161 PRs shipped and the sprint's stated goal (commission the entity library) was not achieved: two entities out of fourteen were meaningfully touched (currency 7/8 done, country 1/6 done). The retrospective note from sprint 19 — "the tooling excuse is spent" — did not hold.
• Sprint planning included 19 stories that were never touched. These functioned as planning inventory, not sprint commitments. Twelve commission entities and five codegen guardrail stories appeared in the sprint table but never had a start date; their presence made the commissioning mission look underway when it was not.
• The codegen model dashboard was a mis-bet: planned, started, and abandoned in the same sprint when the d3-force JS approach proved unworkable at 127 nodes. The sunk cost was modest but the lesson is to time-box explorations explicitly.
• Windows and macOS CI verification slipped for the third consecutive sprint. The root cause (rfl/MSVC C1202 complexity) is cleared, but a story with no start date never gets started.

• Sprint closure now has a formal Postponed section that explicitly separates in-progress stories (where effort was invested) from never-started stories. Story states are normalised at close (BLOCKED→BACKLOG with blockers preserved in Now/Waiting on); the sprint audit catches stale states mechanically.
• Sprint 21 must operate under a hard constraint: no new tooling stories unless they directly unblock a commission task in the same sprint. The commissioning mission is the only permissible mission until the entity library is commissioned.